Simultaneous two-photon spectral and lifetime fluorescence

Simultaneous two-photon spectral and lifetime
fluorescence microscopy
Damian K. Bird, Kevin W. Eliceiri, Ching-Hua Fan, and John G. White
When a fluorescence photon is emitted from a molecule within a living cell it carries a signature that can
potentially identify the molecule and provide information on the microenvironment in which it resides,
thereby providing insights into the physiology of the cell. To unambiguously identify fluorescent probes
and monitor their physiological environment within living specimens by their fluorescent signatures, one
must exploit as much of this information as possible. We describe the development and implementation
of a combined two-photon spectral and lifetime microscope. Fluorescence lifetime images from 16
individual wavelength components of the emission spectrum can be acquired with 10-nm resolution on
a pixel-by-pixel basis. The instrument provides a unique visualization of cellular structures and processes through spectrally and temporally resolved information and may ultimately find applications in
live cell and tissue imaging. © 2004 Optical Society of America
OCIS codes: 180.0180, 180.2520, 170.3650, 170.3890.
1. Introduction
The application of two-photon fluorescence excitation to laser scanning microscopy1 has provided a
powerful tool for three-dimensional imaging of biological specimens. It has several proven advantages over single-photon fluorescence microscopy,
including sample viability,2 an inherent optical
sectioning effect, and superior penetration depth
into dense tissue.3 For these reasons 共and others1兲
two-photon laser scanning microscopy has received
considerable attention as a potential diagnostic imaging modality, primarily by enabling the spatial
distribution of fluorophores 共both endogenous and
exogenous兲 to be visualized from relatively deep
within living tissue.
Recently, optical methods based on fluorescence
spectroscopy have proven to be powerful diagnostic
tools for the study of physiological and morphologWhen this research was performed, the authors were with the
Laboratory for Optical and Computational Instrumentation, Department of Molecular Biology, 271 Animal Sciences, 1675 Observatory Drive, University of Wisconsin-Madison, Madison,
Wisconsin 53706. C.-H. Fan is now with the Department of Mechanical Engineering, Florida International University 10555,
West Flagler Street, EAS 3466, Miami, Florida 33174. The e-mail
address for D. K. Bird is dbird@wisc.edu.
Received 2 March 2004; revised manuscript received 4 May
2004; accepted 18 June 2004.
0003-6935兾04兾275173-10$15.00兾0
© 2004 Optical Society of America
ical changes of tissue at the subcellular level.4 – 8
Analysis of additional fluorescence properties such
as the characteristic emission spectrum and lifetime can provide novel insights into the biophysical
properties of a sample with high molecular specificity and sensitivity.9 –12 The excited-state lifetime of a fluorescent molecule is often of particular
interest because it is both intensity independent
and sensitive to changes in the local microenvironment of the fluorophore. Lifetime determination
may reveal diverse parameters such as pH, molecular
association, oxygen concentration, Ca⫹ concentration 共or other ion concentrations兲, and proteolysis
processing.13,14 To date, the intrinsic features of
fluorescence 共intensity, spectra, and lifetime兲 have
mostly been explored with mutual exclusivity.15–17
Although previous studies utilizing combined spectral and lifetime techniques have been explored,18
in contrast to this report those investigations were
imaged in a different configuration.
In this paper we report on a two-photon fluorescence microscope operating in a simultaneous
spectral–lifetime mode allowing the information
contained within the fluorescence signal to be maximally exploited. We have conducted extensive
characterization experiments on the instrument
and have applied the system to a preliminary analysis of thick tissue sections prepared by staining
protocols that are in common use in pathology laboratories.
20 September 2004 兾 Vol. 43, No. 27 兾 APPLIED OPTICS
5173
Fig. 1. Experimental setup of the combined two-photon spectral
and lifetime microscope. O1, 1.4 NA oil 60⫻ microscope objective;
O2, 0.4 NA 10⫻ microscope objective; D1, D2, dichromatic mirrors;
E, eyepiece; L1, 160-mm transfer lens; BP1, BP2, bandpass filters.
2. System Design
A.
Laser-Scanning Microscope
The experimental arrangement of the new instrument is depicted in Fig. 1. The system is largely
constructed around an inverted microscope 共Nikon,
Diaphot 200兲 and a BioRad MRC-600 confocal scanning head, which houses the optical beam scanning
mechanism and an internal photomultiplier tube
共PMT兲 detector. For source illumination, a Ti:sapphire laser 共Coherent, Mira兲 pumped by an 8-W solidstate laser 共Coherent, Verdi兲 was used to generate
ultrashort optical pulses 共⬃3-nJ pulse energy兲 at a
wavelength of 900 nm and a temporal pulse width of
approximately 120 fs at a repetition rate of 76 MHz.
This beam was coupled to the input laser port of the
MRC-600 scanning head, which was modified with
custom optics to achieve optimum efficiency over the
tuning range of the laser 共approximately 700 –1000
nm兲. In particular, after entering the scanning unit
the pulsed beam encounters a custom dichromatic
mirror 共D1兲 共650DCSP, Chroma Inc.兲 that efficiently
directs approximately 98% of the incident illumination into the scanning mechanism. This contains
two oscillating galvanometer mirrors that generate
the x and y scanning movements.
The scan beam exits the unit through an eyepiece,
which produces a focused scanning spot moving
across the normal focal plane of the imaging objective. The beam is passed to the inverted microscope
through a transfer lens 共L1兲 共01LAO159, MellesGriot兲 and a dichromatic mirror 共D2兲 共650DCSP,
Chroma Inc.兲, which is mounted at 45° in the sliding
filter block of the microscope. This mirror directs
incident illumination up to fill the back aperture of an
imaging objective 共O1兲 关Nikon, 60⫻, 1.4 numerical
aperture 共NA兲, oil兴 and transmits the collected visible
fluorescent light from the sample to a multimode fiber 共core diameter of 1.0 mm, 0.48 NA兲 through a
custom-designed coupling assembly, which contains
5174
APPLIED OPTICS 兾 Vol. 43, No. 27 兾 20 September 2004
a 10⫻, 0.4 NA objective 共O2兲 共Nikon兲 and is directly
attached to the microscope’s bottom-port Keller hole.
The small fraction of fluorescence emission that is not
transmitted by D2 is descanned back to D1 where it is
directed onto the internal PMT of the MRC-600.
This signal is used to construct a two-photon image in
parallel with the spectral and lifetime mode. Fundamental blocking filters, BP1 and BP2 共BG39, Thin
Film Imaging Technologies, Inc.兲 are positioned in
front of the internal PMT and at the proximal end of
the multimode optical fiber in the fiber-optic coupling
assembly. To control the power coupled to the system, a variable neutral-density filter wheel is positioned before the entrance to the MRC-600 scanning
unit. Two-photon images are acquired with custom
software that utilizes a digital signal processor for
generation of the scan waveforms and data acquisition.19 On the basis of axial Rayleigh criterion, the
specifications of O1, and the tuning range of the laser,
the microscope is capable of obtaining optical sections
with a resolution of approximately 1.1–1.5 ␮m.
B.
Combined Spectral and Lifetime Functionality
Measurement of the fluorescence lifetime is commonly achieved either in the frequency or time domain. In the frequency domain the phase shift
between the modulated or pulsed excitation and the
emission from the sample at the fundamental modulation frequency 共or its harmonics兲, together with the
modulation depth, is measured, whereas with timedomain techniques pulsed excitation is used and the
fluorescence decay function is recorded directly.
One increasingly popular time-domain method is
that of time-correlated single-photon counting
共TCSPC兲.20 TCSPC combines a near-ideal utilization of detected photons with extremely high time
resolution, which is limited only by the transit time
spread of the detector. This, together with the fact
that TCSPC can be combined with multiple detectors
for simultaneous multiwavelength detection,21,22
makes the technique particularly attractive for our
application.
To realize our aims, we incorporated a complete
electronic system to record fast light signals by
TCSPC 共SPC-830, Becker & Hickl兲 with our twophoton laser scanning microscope. The SPC-830 is a
compact solution that combines all the components
necessary for TCSPC measurements onto a single
PCI board. Synchronized data collection is achieved
with the SPC-830 board 共which has an internal pixel
clock兲 by use of the x and y laser scanning signals
generated by the digital signal processor for the microscope. Laser pulses are detected by a high-speed
p-i-n photodiode mounted inside the mode-locking
cavity of the laser and are used by the SPC-830 board
to determine the detection time of a photon 共anode
pulse from the PMT兲 relative to the laser pulses.
The SPC-830 unit starts timing on receipt of a detected photon and measures the time lapse until the
next laser pulse. The temporal evolution of the
emission probability after excitation is described by a
histogram of these time spans whereby each counting
event is allocated to a temporal bin for each pixel of
the image. For analysis of the time dependence of
the data it is assumed that the fluorescence decay
histogram for each pixel is well described by a multiexponential decay23:
n
I共t兲 ⫽
兺a
i
exp共⫺t兾␶ i 兲 ⫹ c.
(1)
i⫽0
A Levenberg–Marquardt routine for nonlinear fitting24 is applied to fit a decay curve to the data of this
model function in a separate off-line software package 共SPCImage, Becker & Hickl兲. The SPCImage
software has a deconvolution technique to separate
the decays associated with the different pixels into
the contributions that originate from different emitting species. The final values are approximated by a
convolution of these decays with the recorded instrument response function of the measuring system and
are displayed as a pseudocolored image with color
representing lifetime. The detector employed in our
experiments was a 16-anode photon-counting linear
PMT array 共PML-16, Becker & Hickl兲, which has an
average transient time spread of approximately 200
ps. The multichannel TCSPC technique exploits the
fact that the detection of several photons in different
detector channels in one laser period is unlikely.22
Therefore the single-photon pulses from all detector
channels can be combined into a common photon
pulse line and sent 共together with routing information identifying the PMT channel that generated the
pulse兲 through the normal time measurement procedure of the SPC-830 module, i.e., spectrally resolved
fluorescence lifetime.
Figure 2 shows the physical geometry of the spectrometer implementation. The distal end of the
multimode optical fiber 共FT-1.0-URT, ThorLabs Inc.兲
is mounted in a fiber positioner and delivers the collected fluorescence photons to a sinusoidal concave
diffraction grating G with a focal length f of 10 cm
共Richardson Gratings兲. The NA of the optical fiber
was selected to be comparable with that of the grating
共 f2兲 to minimize loss. The multianode detector is
mounted on an x–y translation stage and positioned
at the principal focus of the diffraction grating where
it detects the fluorescence signal of different wavelength components on the individual cathodes such
that ␭1 ⬍ ␭2 共Fig. 2兲. A pulse generated from the
anode of a PMT within the linear array is amplified,
encoded, and output through a routing channel, C1–
C16, corresponding to the cathode that detected the
fluorescence photon. The fiber, diffraction grating,
and detector are all enclosed within a light-tight box
assembly attached to the optical table to collectively
serve as a 16-channel photon-counting spectrometer.
3. System Characterization
A.
Coupling Efficiency
The geometry of the microscope posed constraints on
the strategies that could be used to determine the
efficiency with which emitted fluorescence from a
Fig. 2. Schematic diagram of the TCSPC spectrometer. Fluorescence collected by the multimode fiber is delivered to the concave diffraction grating G mounted on an x–y translation stage.
The 16-channel detector is positioned at the focus of grating G
where reflected wavelength components impinge on the device
such that ␭1 ⬍ ␭2. Anode pulses are amplified, encoded, and
output through a routing channel C1–C16, corresponding to the
cathode that detected the fluorescence photon.
sample was collected by the multimode fiber. With
the coupling apparatus mounted directly under the
microscope, practical limitations are imposed on the
placement of a powermeter head for a direct measurement of optical power. As a consequence, the
accuracy of the measurement is compromised when
we opt for a less than ideal placement. To address
this issue, we employed a TCSPC technique, which
allows an absolute count of the number of photons
with a direct detection method 共i.e., before entry to
the fiber coupling assembly兲 to be compared with the
total number of photons counted by fiber delivery to
the detector.
A single 10-␮m-diameter fluorescent polymer microsphere 共type YG, Polysciences Inc.兲 with an excitation and emission maxima of approximately 445
and 470 nm, respectively, was used as a sample.
The sphere was isolated from an ensemble of such
spheres dried on a microscope slide and point
scanned with 1.0 mW of 900-nm radiation 共measured
at the back aperture of the imaging objective兲 for 20 s.
After determining the PMT channel that was detecting the peak of the emission spectrum 共i.e., largest
photon counts兲, we executed our photon-counting
software in single-channel mode to build up a single
fluorescence decay histogram in memory corresponding to a single-pixel image. At the completion of the
acquisition, the number of detected photons in each of
the histogram time channels 共bins兲 was summed to
20 September 2004 兾 Vol. 43, No. 27 兾 APPLIED OPTICS
5175
Fig. 3. Calibration of the TCSPC spectrometer. The green 共532nm兲 laser line is positioned on routing channel 4 resulting in the
red 共632.8-nm兲 laser line impinging on channel 14. Linear interpolation reveals a spectral resolution of approximately 10 nm.
give a total count of all photons that impinged on the
detector. This process was then repeated with the
multimode fiber and coupling assembly in place, and
the ratio of the total counts 共the coupling efficiency兲
was measured to be approximately 0.71 ⫾ 0.02 averaged over ten measurements. Each measurement
was obtained from ten different point scan positions
on the microsphere to eliminate the effect of photobleaching on the count measurement. It should be
pointed out that this measurement does not take into
consideration the loss associated with the sinusoidal
diffraction grating, which we measured to be approximately 80%.
B.
Spectrometer Calibration
Calibration of the spectrometer was achieved by standard practice.25 Two laser lines, green and red of
wavelengths 532 and 632.8 nm, respectively, were
coupled to the multimode fiber by a 10⫻, 0.25 NA
objective 共O2 in Fig. 1兲 and delivered to the diffraction
grating G 共Fig. 2兲. With the NA of the fiber being
comparable to that of the diffraction grating, the
emerging beams approximately filled the diameter of
the concave grating. Fine adjustment of the x and y
translation stages upon which the grating was
mounted allowed maximum efficiency to be attained.
The fine position of the 16-channel detector placed at
the first-order focal plane was controlled with an
identical arrangement of stages. Figure 3 shows the
position of each of these wavelengths on a given routing channel. For calibration, we positioned the
532-nm line on routing channel 4 by translating the
stage of the detector. It can be seen from Fig. 3 that
this resulted in the illumination of 632.8 nm to impinge on routing channel 14, which by a process of
linear interpolation corresponds to a spectral resolution of approximately 10 nm. It is interesting to
note the approximately equal photon counts in each
5176
APPLIED OPTICS 兾 Vol. 43, No. 27 兾 20 September 2004
Fig. 4. Second-harmonic spectra recorded for incident illumination wavelengths of 共a兲 900 nm and 共b兲 860 nm. The harmonic
peaks are at 450 and 430 nm, respectively.
time bin of the spectral channels 4 and 14. This
phenomenon can be attributed to the continuous
wavelength nature of the laser source because, under
this condition, there is no correlation of the detection
of a photon relative to an excitation pulse.
As a further measure of the spectrometer performance, we measured the second-harmonic spectrum generated from a nonlinear ␤-bariumborate
共BBO兲 crystal for a range of illumination wavelengths. Figures 4共a兲 and 4共b兲 show the measured
spectra for two illumination wavelengths, 900 and
860 nm, respectively. In this arrangement the
PMT array was first translated such that continuous wavelength illumination of 532 nm impinged on
routing channel 16, allowing the detection of spectral components as short as approximately 382 nm
on routing channel 1. It can be clearly seen from
Figs. 4共a兲 and 4共b兲 that the second-harmonic generation 共SHG兲 spectra are accurately resolved on routing channels 8 共450 nm兲 and 6 共430 nm兲 and their
SHG maxima coincide with the peak of the excitation wavelength. The slight photon spill into adjacent PMT channels 共an effect not observed in Fig.
3兲 is due to the fact that the transform-limited ul-
instrument is to be applied to applications that require a more quantitative measurement 共as opposed
to our relative analysis兲, the spectral dependence of
the radiant sensitivity of the multianode PMT assembly 共R5900U-L16, Hamamatsu兲 is known26 and can
be employed to address this issue.
C.
Fig. 5. System response function of the combined spectral–
lifetime microscope to a SHG signal from a BBO crystal measured
at routing channel 8 of the multianode photon-counting PMT.
The FWHM of the curve is approximately 295 ps. Inset shows the
measured system response at FWHM as a function of anode number in the PMT array.
trashort pulses of the excitation source have a bandwidth of approximately 16 nm.
It should be pointed out that the variation in the
spectral response of any given cathode in the linear
PMT array was not taken into consideration in the
calibration process. For experiments in which the
System Response
The system response is a critical parameter when
fluorescence lifetime data are modeled because it is
convolved with the time slope of a 共multi兲exponential
decay.22 A fast conventional PMT may have a transit time spread as short as approximately 150 ps.
For a microchannel plate PMT, this time can be a
short as 25–30 ps. Although much shorter lifetimes
can be determined by deconvolution of the recorded
exponential decay function21 in principle, it is the
response of the PMT that fundamentally limits the
time resolution of the system in practice. The
Becker & Hickl PML-16 detector used in our experiments has a quoted response time of approximately
200 ps. This of course represents an ideal situation
and does not take into consideration systemdependent optical effects. There are numerous factors, including light scattering and wavelengthdependent light path differences that will in general
cause degradation of the system response. This is
particularly applicable when a multimode optical fiber is used to deliver the fluorescence photons to a
Fig. 6. Single-channel control lifetime image of a 10-␮m-diameter fluorescent polymer microsphere imaged at 920 nm showing an
approximately uniform 2.28 ⫾ 0.1 ns lifetime distribution across the field.
20 September 2004 兾 Vol. 43, No. 27 兾 APPLIED OPTICS
5177
The results are given in Fig. 5, which depicts the
number of measured SHG photons by routing channel 8 of the PML-16 detector as a function of time.
The measured full width at half-maximum 共FWHM兲
of this curve is approximately 295 ps, which is increased by approximately 95 ps compared with the
manufacturer specifications. This increase is
largely attributable to dispersion in the multimode
optical fiber. As a further test, the inset in Fig. 5
shows the measured system response at FWHM as a
function of the anode number in the linear PMT array, where it is clear that any given channel of the
detector has a comparable system response. This
result is of important physical significance and demonstrates the ability of the instrument to resolve fluorescence lifetime measurements uniformly across
the spectrum.
4. Experimental Results
A.
Fig. 7. 共a兲 Recorded emission spectrum and 共b兲 fluorescence lifetimes from a sample comprising a mix of two distinct species of
fluorescent polymer microspheres with overlapping emission spectra imaged at 920 nm. Two peaks are observable in 共a兲 at approximately 460 and 570 nm. Lifetime analysis of these wavelengths
in 共b兲 reveals two characteristic lifetimes, which can be used to
identify the origin of any given wavelength component of the spectrum.
remote detector due to multipath propagation within
the fiber core.
To characterize the performance of the new instrument, we measured the system response to the SHG
signal from a nonlinear BBO optical crystal. This is
a particularly attractive calibration technique because 共by virtue of the SHG process兲 the signal has a
zero lifetime and a narrow-bandwidth spectrum
关Figs. 4共a兲 and 4共b兲兴. The BBO crystal combines a
number of unique features including wide transparency and phase-matching ranges, a large nonlinear
coefficient, high damage threshold, and excellent optical homogeneity. When the system response is
monitored in real time with the SPC-830 board running in oscilloscope mode and in the absence of
background fluorescence processes, the optical configuration and alignment can be optimized to yield
the fastest response function.
5178
APPLIED OPTICS 兾 Vol. 43, No. 27 兾 20 September 2004
Wavelength-Resolved Fluorescence Lifetime
As an initial control, we measured the variation in
lifetime across a fluorescent polymer microsphere of
10 ␮m in diameter 共type YG, Polysciences Inc.兲.
This allowed us to access the imaging performance
and lifetime uniformity across the optical field because the microspheres contain a single species of
fluorescent dye molecule. The mean lifetime image
is given in Fig. 6 and indicates an almost uniform
distribution of monoexponential decay times within
the microsphere of approximately 2.28 ⫾ 0.1 ns.
This result is in good agreement with earlier measurements of similar microspheres that varied only
by virtue of their diameter.27
To demonstrate the ability of the new instrument
to resolve particular wavelength components by the
characteristic fluorescence lifetime, we simultaneously imaged and recorded, by two-photon excitation at 920 nm, the lifetime of two distinct species of
fluorescent polymer microspheres. Microspheres
共Polysciences Inc.兲 of type YG 共Fluorescein兲 and YO
共Rhodamine兲 were specifically selected because of
clear overlap in their emission spectrum. We produced a sample of randomly dispersed spheres by
drying a drop of each solution in which the microspheres were suspended onto a microscope slide.
The recorded emission spectrum from the sample is
shown in Fig. 7共a兲 and is not representative of any
given pixel, but rather results from photon counts
that occurred as a consequence of our scanning the
entire image for approximately 10 s. In this arrangement, the spectrometer was calibrated such
that the 532-nm laser line impinged on routing channel 10.
Observation of Fig. 7共a兲 reveals two peaks. The
first occurs at approximately 460 nm and the second
appears at approximately 570 nm, corresponding to
routing channels 3 and 14, respectively. Because
the peak emission wavelength of the YO microsphere
is approximately 570 nm and the tail of the YG emission spectra extends beyond 600 nm,28 it is difficult to
conclude from this measurement if the 570-nm wave-
Fig. 8. Two-photon fluorescence intensity images of a transverse section though the medulla of a Cynomolgus monkey kidney acquired
at the 共a兲 480-, 共b兲 510-, 共c兲 550-, and 共d兲 580-nm wavelength components of the emission spectrum. The sample was stained with the
methyl green fluorochrome and imaged with an illumination wavelength of 920 nm.
length component is a result of the Fluorescein or the
Rhodamine fluorophore. However, analysis of the
fluorescence lifetimes of individual spectral components can provide further insight. Figure 7共b兲 is the
result of our applying a Levenberg–Marquardt nonlinear fitting routine to the accumulated histogram
data in all time bins of routing channels 3 and 14.
As is shown in Fig. 7共b兲 the fits clearly reveal two
independent monoexponential decays having lifetimes of approximately 2.31 ⫾ 0.1 ns 共curve A兲 and
3.4 ⫾ 0.1 ns 共curve B兲, respectively, which correspond
to the 460- and 570-nm wavelength components of
the emission spectrum. It is clear from this result
that the 570-nm spectral component is a consequence
of the YO microspheres because the shorter wavelength component lifetime is in good agreement with
that measured by our control experiment where only
the YG microspheres were considered. It should
also be pointed out that each peak in the emission
spectrum is within approximately 1.0% of those
quoted by the manufacturer of the microspheres.28
B.
Combined Two-Photon Spectral–Lifetime Imaging
To explore the potential of the new simultaneous twophoton spectral–lifetime imaging system as a diagnostic tool for pathological investigations, we
initiated preliminary imaging studies of a thick primate histological specimen 共Cynomolgus monkey kidney兲. Tissue was excised and fixed in neutral
buffered formalyne and sectioned with a vibratome in
200-␮m sections. The sample was then immersed in
0.5% methyl green 共in phosphate buffer兲 and taken
through 95% and 100% ethanol. Finally, the tissue
is cleared in xylene and mounted with cytosol 60.
Figures 8共a兲– 8共d兲 depict four two-photon fluorescence
intensity images of a transverse section of medulla
obtained from the Cynomolgus monkey kidney acquired at four representative wavelength components of the emission spectrum: 480, 510, 550, and
580 nm. For clarity, each image is depicted so that
it appears to be the actual color of the spectral component. The total acquisition time for the entire
20 September 2004 兾 Vol. 43, No. 27 兾 APPLIED OPTICS
5179
Fig. 9. Fluorescence lifetime images of a transverse section though the medulla of a Cynomolgus monkey kidney acquired at the 共a兲 480-,
共b兲 510-, 共c兲 550-, and 共d兲 580-nm wavelength components of the emission spectrum. The sample was stained with the methyl green
fluorochrome and imaged with an illumination wavelength of 920 nm.
Fig. 10. Multiexponential best-fit analysis of a representative
pixel 共74, 65兲 for the 共a兲 480-, 共b兲 510-, 共c兲 550-, and 共d兲 580-nm
wavelength components of the emission spectrum.
5180
APPLIED OPTICS 兾 Vol. 43, No. 27 兾 20 September 2004
Fig. 11. Fluorescence lifetime analysis across the 450 – 600-nm
spectral range for a representative pixel 共74, 65兲 on each lifetime
image. The dotted and dashed curves depict the variation in the
short 共␶1兲 and long 共␶2兲 lifetime components as a function of wavelength, respectively. The ratio of amplitudes 共a1兾a2兲 as a function
of wavelength 共solid curve兲 reveals the relative contribution of
these components in the multiexponential sum.
data set was approximately 3 min with an average
photon count rate of approximately 7.5 ⫻ 105 photons兾s. The focused spot size and the average illumination power at the sample were approximately
0.8 ␮m and 4.4 ⫾ 0.2 mW, respectively. It is clear
that any given image corresponding to a particular
wavelength component of the emission spectrum reveals particular features about the sample that are
not necessarily discernable by other wavelength components. This is particularly evident in Fig. 8共a兲
where various nuclei that are well defined in Fig. 8共d兲
are not easily identifiable. Figures 9共a兲–9共d兲 illustrate pseudocolored fluorescence lifetime images corresponding to each of the four representative
wavelength components given in Figs. 8共a兲– 8共d兲.
The entire data set 共which occupies approximately 32
Mbytes on the disk兲 comprises 16, 128 ⫻ 128 pixel
images, with each pixel being a lifetime photon count
histogram comprising 64 time bins that can each contain up to 216 photon counts.
Visual analysis of Figs. 9共a兲–9共d兲 reveals discrete
lifetime distributions within the sample that correspond to signals emanating from different structures,
which likely arise from the varying microenvironments in different regions of the cell or tissue. General observation of all four images reveals that the
nuclei in the collecting ducts 共identified by their columnar epithelium兲 and collecting tubules 共which
have a cuboidal epithelium兲 have shorter lifetimes
than that of the surrounding cytoplasm. However,
examination of the image data with greater scrutiny
on a pixel-by-pixel basis over the representative fluorescence spectrum components reveals other subtle
changes that may be of great characteristic significance.
By way of example, we considered the same pixel
共74, 65兲 on each spectral–lifetime image given in Figs.
9共a兲–9共d兲. Figure 10, curves 共a兲–共d兲, depict the
best-fit analysis of the photon histograms for our chosen representative pixel on each of these spectral–
lifetime images 关Figs. 9共a兲–9共d兲兴. It is clear from
examination of the 1兾e point in Fig. 10 that each of
these points has a different fluorescence lifetime.
Furthermore, it is interesting to note that these data
were best fitted with two exponential components in
the intensity sum 关Eq. 共1兲兴. When several distinct
decay rates from a single fluorophore are observed
within a sample or as we demonstrate in Figs. 9共a兲–
9共d兲, it is likely that the lifetime of the fluorophore is
being locally modified by the close association of a
molecule共s兲 within the specimen that alters the fluorescent lifetime by resonant energy transfer. In this
case one can extract the respective lifetimes ␶i and
the distribution of the respective coefficients ai from
the model function and analyze the ratio of the associated and unassociated fluorophore at every pixel in
the image.
Employing this scheme reveals an approximately
equally weighted distribution of short 共␶1, approximately 500 ps兲 and long 共␶2, approximately 1900 ps兲
lifetime components for the 480-nm spectral channel,
whereas at the longer 580-nm wavelength component
the short component ␶1 remains approximately unchanged 共520 ps兲 and the long component ␶2 increases
to approximately 3200 ps. Furthermore, at this
spectral component the distribution of these lifetimes
is significantly altered by approximately 12% with
the longer lifetime ␶2 being the dominant component
in the exponential sum. Analysis across the entire
fluorescence spectrum is provided in Fig. 11 where
the short lifetime component ␶1 共dotted curve兲, the
long lifetime component ␶2 共dashed curve兲, and the
ratio of the amplitude coefficients a1兾a2 共solid curve兲
are plotted as a function of wavelength. Although
on first observation it appears that little in the way of
change is occurring below a wavelength of approximately 510 nm when one considers only the value of
either ␶1 or ␶2, observation of the relative contributions of each of these components to the exponential
decay a1兾a2 共solid curve兲 reveals the contrary. It is
also interesting to note that the general trend of the
data tends to suggest an increase in the fluorescence
lifetime as the wavelength in the fluorescence spectrum increases.
5. Conclusion
We have reported on the design and implementation of a combined two-photon spectral and lifetime
microscope that incorporates a linear array of 16
single-photon-counting detectors for simultaneous
measurement of two-photon-excited fluorescence
emission spectra and lifetime on a pixel-by-pixel
basis. The new system is capable of measuring a
160-nm spectral window within the range of approximately 380 – 680 nm 共by recalibration兲 with
approximately 10.0-nm resolution, together with
the excited-state lifetime of the fluorophore for each
individual wavelength component with a time resolution of approximately 295 ps. The typical acquisition time of an entire spectral–lifetime image
set is approximately 3 min, which is required to
record a sufficient number of photon counts for precise curve fitting. This limitation can be reduced
by one counting fewer photons; however, this comes
at the expense of accuracy. We have performed
extensive characterization of the system using a
variety of fluorescent samples, and it has proven to
be effective to resolve overlapping spectra by analysis of the characteristic lifetimes. Combined
spectral–lifetime analysis may be particularly powerful to analyze studies involving fluorescent resonant energy transfer. Measurements of the
reduction of donor lifetime together with a change
in intensity of donor and acceptor wavelengths during a fluorescent resonant energy transfer interaction may avoid some of the ambiguities due to
unknown concentration changes inherent in measurements that involve only relative changes in intensity of the donor and acceptor wavelengths.
We have begun to explore the feasibility of a combined spectral–lifetime imaging mode as a useful tool
for pathological analysis through systematic analytic
evaluation of spectral–lifetime data sets obtained
from a histological tissue sample. The new instru20 September 2004 兾 Vol. 43, No. 27 兾 APPLIED OPTICS
5181
ment allows the information contained within the
fluorescence signal to be maximally exploited and
may provide new levels of diagnostically valuable information for identification of subtle characteristic
properties of cells and tissue. Although the analysis
presented in this paper is by no means an exhaustive
evaluation of the image data, it does reveal the potential analytical power of a combined spectral and
lifetime imaging modality. It also brings to fruition
some of the more practical challenges that need to be
addressed for the analysis of such complex data sets.
Future goals include use of the instrument for various in vivo biological studies.
This research is supported under the National
Institutes of Health, National Institute of Biomedical Imaging and Bioengineering grant R01
EB000184. The authors thank Gary Lyons for histological identification; Al Kutchera of Midwest Microtech for the histology samples; Muhammad
Nazir, Vikash Agarwal, and Ying Rao for systemspecific software development; and Gary Hammersley and John Peterson for the manufacturing of
custom system components.
References
1. W. Denk, J. H. Strickler, and W. W. Webb, “Two-photon laser
scanning fluorescence microscopy,” Science 248, 73–76 共1990兲.
2. J. M. Squirrell, D. L. Wokosin, J. G. White, and B. D. Bavister,
“Long-term two-photon fluorescence imaging of mammalian
embryos without compromising viability,” Nat. Biotechnol. 17,
763–767 共1999兲.
3. V. E. Centonze and J. G. White, “Multiphoton excitation provides optical sections from deeper within scattering specimens
than confocal imaging,” Biophys. J. 75, 2015–2024 共1998兲.
4. X. F. Wang, A. Periasamy, and B. Herman, “Fluorescence lifetime imaging microscopy 共FLIM兲: instrumentation and applications,” Crit. Rev. Anal. Chem. 23, 365–369 共1992兲.
5. T. W. J. Gadella, T. M. Jovin, and R. M. Clegg, “Fluorescence
lifetime imaging microscopy 共FLIM兲: spatial resolution of microstructures on the nanosecond time scale,” Biophys. Chem.
48, 221–239 共1993兲.
6. M. J. Cole, J. Siegel, S. E. D. Webb, R. Jones, K. Dowling,
M. J. Dayel, D. Parsons-Karavassilis, P. M. W. French, M. J.
Lever, L. O. D. Sucharov, M. A. A. Neil, R. Juskaitis, and T.
Wilson, “Time-domain whole-field fluorescence lifetime imaging with optical sectioning,” J. Microsc. 共Oxford兲 203, 246 –
257 共2001兲.
7. H. C. Gerritsen and K. deGrauw, “One and two-photon confocal
fluorescence lifetime imaging and its applications,” in Methods
in Cellular Imaging, A. Periasamy, ed. 共Oxford U. Press, New
York, 2001兲, pp. 309 –323.
8. J. R. Lakowicz, Principles of Fluorescence Spectroscopy 共Plenum, New York, 1999兲.
9. G. A. Wagnieres, W. M. Star, and B. C. Wilson, “In vivo fluorescence spectroscopy and imaging for oncological applications,” Photochem. Photobiol. 68, 603– 632 共1998兲.
10. P. J. Tadrous, J. Siegel, P. M. French, S. Shousha, el-N. Lalani,
5182
APPLIED OPTICS 兾 Vol. 43, No. 27 兾 20 September 2004
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
and G. W. Stamp, “Fluorescence lifetime imaging of unstained
tissues: early results in human breast cancer,” J. Pathol. 199,
309 –317 共2003兲.
M. E. Dickinson, E. Simbuerger, B. Zimmermann, C. W. Waters, and S. E. Fraser, “Multiphoton excitation spectra in biological samples,” J. Biomed. Opt. 8, 329 –338 共2003兲.
M. E. Dickinson, G. Bearman, S. Tille, R. Lansford, and S. E.
Fraser, “Multi-spectral imaging and linear unmixing add a
whole new dimension to laser scanning fluorescence microscopy,” BioTechniques 31, 1272–1278 共2001兲.
J. R. Lakowicz, H. Szmacinski, K. Nowaczyk, K. W. Berndt,
and M. Johnson, “Fluorescence lifetime imaging,” Anal. Biochem. 202, 316 –330 共1992兲.
P. I. Bastiaens and A. Squire, “Fluorescence lifetime imaging
microscopy: spatial resolution of biochemical processes in the
cell,” Trends Cell Biol. 9, 48 –52 共1999兲.
W. Wang, J. B. Wyckoff, V. C. Frohlich, Y. Oleynikov, S.
Huttelmaier, J. Zavadil, L. Cermak, E. P. Bottinger, R. H.
Singer, J. G. White, J. E. Segall, and J. S. Condeelis, “Single
cell behavior in metastatic primary mammary tumors correlated with gene expression patterns revealed by molecular
profiling,” Cancer Res. 62, 6278 – 6288 共2002兲.
N. Ramanujam, “Fluorescence spectroscopy of neoplastic and
non-neoplastic tissues,” Neoplasia 2, 89 –117 共2000兲.
K. W. Eliceiri, C. H. Fan, G. A. Lyons, and J. G. White, “Analysis of histology specimens using lifetime multiphoton microscopy,” J. Biomed. Opt. 8, 376 –380 共2003兲.
Q. S. Hanley, D. J. Arndt-Jovin, and T. M. Jovin, “Spectrally
resolved fluorescence lifetime imaging microscopy,” Appl.
Spectrosc. 56, 155–166 共2002兲.
K. W. Eliceiri, “WiscScan: A DSP based Acquisition System
for Laser Scanning Microscopes,” http:兾兾www.loci.wisc.edu兾
WiscScan兾.
D. V. O’Connor and D. Phillips, Time Correlated Single Photon
Counting 共Academic, London, 1984兲.
W. Becker, A. Bergmann, C. Biskup, T. Zimmer, N. Klöcker,
and K. Benndorf, “Multiwavelength TCSPC lifetime imaging,”
in Multiphoton Microscopy in the Biomedical Sciences II, A.
Periasamy and P. T. C. So, eds., Proc. SPIE 4620, 74 – 84
共2002兲.
W. Becker, A. Bergmann, M. A. Hink, K. König, K. Benndorf,
and C. Biskup, “Fluorescence lifetime imaging by timecorrelated single-photon counting,” Microsc. Res. Tech. 63,
58 – 66 共2004兲.
A. Schönle, M. Glatz, and S. W. Hell, “Four-dimensional multiphoton microscopy with time-correlated single-photon counting,” Appl. Opt. 39, 6306 – 6311 共2000兲.
W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling, Numerical Recipes in C, 2nd ed. 共Cambridge U. Press,
Cambridge, UK, 1993兲.
E. Hecht, Optics, 2nd ed. 共Addison-Wesley, Reading, Mass., 1990兲.
Hamamatsu Photonics K. K. Electron Tube Center, R5900UL16 Series Data Sheet, http:兾兾www.hamamatsu.com
共Hamamatsu Photonics, Hamamatsu City, Japan, 2003兲.
M. Straub and S. W. Hell, “Fluorescence lifetime threedimensional microscopy with picosecond precision using a
multifocal multiphoton microscope,” Appl. Phys. Lett. 73,
1769 –1771 共1998兲.
Fluoresbrite Microparticles, Technical Data Sheet 431 共Polysciences Inc., Warrington, Pa., 2001兲.