Open Access
Add to BibSonomyAbstract
When performing multiphoton fluorescence lifetime imaging in multiple spectral emission channels, an instrument response function must be acquired in each channel if accurate measurements of complex fluorescence decays are to be performed. Although this can be achieved using the reference reconvolution technique, it is difficult to identify suitable fluorophores with a mono-exponential fluorescence decay across a broad emission spectrum. We present a solution to this problem by measuring the IRF using the ultrafast luminescence from gold nanorods. We show that ultrafast gold nanorod luminescence allows the IRF to be directly obtained in multiple spectral channels simultaneously across a wide spectral range. We validate this approach by presenting an analysis of multispectral autofluorescence FLIM data obtained from human skin ex vivo.
©2011 Optical Society of America
1. Introduction
Multiphoton fluorescence microscopy has become an established technique for optically sectioned subcellular imaging in optically scattering media [1]. The well-known advantages of this technique over confocal microscopy include deeper penetration into tissue and the lack of a confocal pinhole, which allows scattered fluorescence photons to be collected, and the ability to excite fluorophores whose single-photon excitation lies in the ultraviolet. Multiphoton microscopy has found numerous applications in tissue imaging, including in vivo imaging of human skin [2,3], and can be readily combined with time-correlated single photon counting (TCSPC) detectors to realize fluorescence lifetime imaging (FLIM) [2,4]. Fluorescence lifetime measurements of autofluorescent compounds have been shown to provide an indicator of disease [5–9] and therefore the combination of multiphoton excitation and FLIM offers the potential for obtaining label free diagnostic information that can be measured in vivo [2].
Multiphoton fluorescence microscopy can be extended by resolving the fluorescence emission spectrum, which is particularly advantageous for autofluorescence applications where the signal originates from several different molecules with different spectral emission properties [10] and it is also possible to combine spectral and temporal measurements of the fluorescence signal, e.g., [11]. Reliable and accurate fitting of multispectral autofluorescence decay data would offer valuable diagnostic potential and would facilitate reliable, quantitative and repeatable analysis. This in turn would aid the comparison of independent fluorescence lifetime measurements at different laboratories, providing more robust interpretation of results and would facilitate the machine-assisted interpretation of FLIM results that is likely to be required for clinical microscopy systems.
In order to accurately fit complex fluorescence decays, one must perform a measurement of the system’s temporal instrument response function (IRF). This information is then used during data analysis to compensate for distortions in the data caused by the IRF. The ideal way to determine the IRF is to directly measure it using a sample with an extremely fast response. In two-photon microscopy, the reported methods to directly record the IRF include the use of the hyper-Rayleigh signal from gold nanospheres [12] or second harmonic generation (SHG) in, e.g., urea or potassium diphosphate crystals [13]. However, in both cases, the signal is produced at half the excitation wavelength. Ideally, the instrument response function should be acquired with identical instrument settings and emission filters as the acquired fluorescence image data. For single channel FLIM, it may be satisfactory to measure the IRF using a SHG sample if the emission filter allows the collection of the SHG signal. However, if a narrower more specific emission filter is employed, then this will often block the SHG signal and the emission filter must be changed for the IRF measurement. For spectrally resolved FLIM [11,14,15], it is more difficult to use SHG to measure the IRF. If the system is designed with dichroic beam splitters and multiple detectors, as is the case for the system presented below, then the dichroics must be changed to direct the SHG light into each channel for measurement of the IRF. If a spectrograph based multispectral FLIM system with multi-anode PMT is employed, e.g., [11], then the spectrograph grating must be scanned so that SHG signal falls on each anode sequentially, or the system must be rearranged in some other way to allow the SHG signal to be incident on each of the detector elements. This experimental complexity increases with the number of detection channels. A hyper-Rayleigh or SHG scattering signal is therefore often not a convenient method for acquiring the IRF when performing multispectral FLIM measurements.
A further challenge for IRF measurements is that the temporal response of the photodetector(s) may have a spectral dependence, known as the “color shift”. Such dependencies may arise from the wavelength dependence of the energy of the initial photoelectron, or from a chromatic bias in the angle of incidence of the detected photon or the area of detection on the photocathode [16]. These effects can lead to a temporal shift or a broadening of the IRF that is dependent on wavelength. Although this is not a major issue for MCP-PMTs or many contemporary PMTs [16,17], it is still an issue with avalanche photodiodes [16,18]. When using such detectors, the IRF should therefore be acquired over the same spectral window as that used for recording the signal from the sample under study, but this is not possible using a hyper-Rayleigh or SHG scattering signal.
One approach to address these challenges is to measure the IRF using a fluorescent dye with a lifetime that is so short that it can be neglected in the subsequent analysis, i.e. the decay can be treated as an impulse response, with the fluorescent dye being chosen such that it emits in the same wavelength range as the sample being studied. However, it is not always possible to find such short lifetime dyes for every spectral window of interest. For example, the emission spectra of short lifetime fluorescent dyes reported in [19] only extend upwards from ~525 nm and so do not provide full coverage of the visible spectral range. If the dye lifetime is not sufficiently short, this approach can compromise the analysis of multiexponential decays with short lifetime components, e.g. from melanin and NAD(P)H [20,21].
An alternative approach to measuring the IRF is to measure the response of the instrument to a reference fluorophore with a known, single exponential lifetime and then to modify the fitting model. Full details of this “reference reconvolution” technique can be found in [22]. In conventional reconvolution, the measured IRF is convolved with the fluorescence decay model, e.g. a sum of one or more exponential decays, to produce a curve that can be used by the fitting algorithm. The fitting algorithm adjusts the parameters of the decay model until the best fit to the data is obtained, e.g. via a non-linear least-squares (NLLSQ) fitting algorithm. If a single exponential reference decay is measured instead of the IRF, then the fluorescence decay model must be modified by deconvolving it by a single exponential decay with the lifetime of the reference fluorophore. Then, convolution of the reference decay with the modified fitting model produces the desired curve for decay fitting. Using this technique increases the range of available fluorescent reference samples to include those with longer lifetimes and therefore it is more likely that a reference fluorophore can be identified corresponding to the required spectral emission. However, it is still difficult to identify suitable dyes that have monoexponential decays over a wide spectral emission range. In addition, preparation of the dyes must be performed under carefully controlled conditions in order to ensure a monoexponential fluorescence decay profile with the correct lifetime [23] and furthermore, the dyes will have a limited shelf life.
We propose the use of gold nanorods as an alternative method to directly measure the IRF across a broad spectral range for two-photon excited FLIM. Luminescence from bulk gold was first reported by Mooridan in 1969 [24] with a very low quantum yield estimated to be 10−10. The origin of the luminescence occurs following the creation of an electron-hole (e-h) pair in the 5d and 6sp bands of bulk gold. The e-h pair has a very fast non-radiative decay rate due to Coulomb scattering, which results in a very low quantum yield and an ultrafast decay. However, a dramatic increase in the quantum yield to ~10−4 is observed with gold nanorods [25]. It is thought that this increase is due the availability of a plasmon resonance coinciding with the difference between d-band and sp-band energy levels, i.e. plasmons are resonantly emitted by the recombination of d-band holes and sp-band electrons and can then decay to emit a photon [26,27]. Such a process would increase the radiative decay rate and therefore should increase the quantum yield and decrease the decay time of the luminescence if the non-radiative decay rate does not change significantly. Temporally resolved measurements of plasmon enhanced gold luminescence have shown the ultrafast decay to be much faster than the instrument response function of the equipment used [26,28–30], which was < 50 fs in [31]. Longer components of the decay (~ps) have also been observed [29–31], however the relative contribution of these long components are very low and has been shown to vary with the excitation pulse energy [29]. Much longer decay components have been reported [32], however, in the data presented in this paper, they are not observed. The multi-photon excitation of the luminescence is also greatly enhanced by the presence of the plasmon oscillation [33] whose broad resonance wavelength band depends on the statistical distribution of aspect ratios of the nanorod ensemble [25,34,35]. Hence there is enhancement of both the excitation and emission efficiencies of multi-photon excited luminescence from gold nanorods [36]. The ultrashort lifetime of gold nanorod luminescence [31] together with their broad emission spectrum that extends across the visible region [24,25,36], makes them ideal candidates as sources for IRF measurements in multiphoton fluorescence lifetime experiments – particularly for multispectral instruments. We note that the ultrashort fluorescence lifetime of gold nanorods has been exploited previously in multiphoton FLIM measurements as a contrast agent that is easily distinguished from other fluorophores [37].
In this paper, we introduce a custom-built TCSPC detector with four spectral channels that has been integrated into the commercially available two photon DermaInspect® system (JenLab) for ex vivo and in vivo multiphoton FLIM of skin. We then present luminescence measurements of gold nanorods and demonstrate their use for the direct and simultaneous measurement of the IRF in all four channels of the multispectral detector. This measurement is carried out in a single acquisition step using a physically robust sample of gold nanorods embedded in polyvinyl alcohol and is therefore directly applicable for use in a clinical setting. Finally, we use the IRF data acquired using the gold nanorods to analyze experimental multiphoton excited FLIM data acquired from ex vivo skin tissue. We note that although this work was carried out with a specific instrument and application in mind, the techniques is applicable and relevant to all current implementations of multiphoton FLIM including live cell imaging, FLIM-FRET measurements and imaging of disease models in live animals.
2. Materials and methods
2.1 Gold nanorods
All chemicals were purchased from Sigma-Aldrich Inc., USA, and used as received. Films of gold nanorods and polyvinyl alcohol were prepared by modification of the procedures described in [38–40]. Aqueous suspensions of cetrimonium-stabilized gold nanorods were first synthesized as in [41], i.e. by two-step overgrowth and elongation of an initial population of gold nanospheres through reduction of chloroauric acid by ascorbic acid in the presence of cetrimonium bromide and silver nitrate. The particles realized for these experiments consisted of a majority of rods (~96% by volume) with a prevalent dog-bone profile [42,43] and a minority of by-products (~4% by volume), which displayed a variety of shapes with some preference for cubes. Transmission electron micrographs of these particles revealed a broad distribution of lengths (29 ± 6 nm) and diameters (7.2 ± 1.5 nm), with a volume weighted aspect ratio of 4.3 ± 0.7. At least two days after the synthesis, particles were purified by cycles of centrifugation and decantation in order to remove the excess of cetrimonium bromide and unreacted chemicals and finally re-dispersed and homogenized into aqueous polyvinyl alcohol (PVA, MW ~100000) (10% w/w) at a rate of 400 µg.ml−1. 50 µl of these mixtures were poured into polystyrene molds and left to dry overnight in order to obtain hybrid films with a thickness ~200 µm, which were stored in methanol until utilization. For microscopy, the nanorod PVA film was placed on a slide without a mounting medium and a cover slip was placed on top. We note that these samples can be produced at low cost and can be used repeatedly. The extinction spectrum of the nanorods was measured by placing the slide in the beam path of a spectrophotometer (UV3101, Shimadzu).
2.2 Instrumentation
All fluorescence measurements were performed using a commercially available 2-photon tomograph designed for in vivo skin imaging (DermaInspect®, JenLab GmbH). A description of the DermaInspect® can be found in [2]. Briefly, IR excitation is provided by a tunable Ti:sapphire laser (MaiTai DeepSee, SpectraPhysics) that is directed onto the specimen via a galvo-scanner and a 1.3 NA oil immersion 40× objective. The collected fluorescence can then be directed to an intensity PMT, to our custom-built 4-channel TCSPC detector or to our custom-built spectrometer as shown in Fig. 1(a) . A full description of our custom-built detector system and its design can be found in [44]. The custom-built spectrometer is based on an optical fiber-bundle spot-to-line converter, prism and CCD camera arranged as shown in Fig. 1(b). Wavelength calibration of the spectrometer was carried out using the second harmonic signal from urea crystals excited by the MaiTai at different excitation wavelengths. The excitation wavelength was independently measured using a spectrometer (HR2000, Ocean Optics). Intensity calibration of the spectrometer was carried out using a calibrated white light source (LS-1-CAL, Ocean Optics). The 4-channel TCSPC detector utilizes an arrangement of dichroics and filters to spectrally separate the fluorescence onto four PMTs (H7422P-40, Hamamatsu Photonics) as shown in Fig. 1(c). The PMTs were connected to an SPC830 (Becker & Hickl, Berlin) card to perform time correlated single photon counting. The spectral resolution of the 4 channel detector (Fig. 1(d)) was calculated using the transmission curves of the dichroics and filters as provided by the manufacturer (Semrock, IL, USA).
Fig. 1 Showing a schematic of the DermaInspect and the detectors. (a) shows a simplified diagram of the excitation path. The DermaInspect has a slider in the emission path which allows the collected fluorescence to be directed to one of three different detectors: an intensity PMT, the spectrometer (b) or the 4-channel TCSPC detector (c). The spectral response of the 4-channel detector is shown in (d).
In order to maximize the amount of fluorescence reaching the detectors, a ray tracing algorithm was employed when designing the custom-built 4-channel detector. Through proper choice of lenses and their positioning, as well as the positioning of every aperture in the collection beam path, the geometric transmission efficiency of the 4-channel detector was >85% across two thirds of the field of view of the objective. In addition to maximizing the field of view of the 4-channel detector, it also maximizes the collection of scattered light, which in turn allows deeper imaging into tissue.
Measurement of the excitation power dependence of the emission intensity of the gold nanorods was performed using a Leica TCS SP5 FLIM microscope. Multiphoton excitation was provided by a Ti:sapphire laser (MaiTai, SpectraPhysics) and the specimen was illuminated via a 0.75 NA, 40 × air objective. The excitation power was measured at the output of the objective. The fluorescence was directed onto a single photon counting detector connected to a SPC830 (Becker & Hickl) card. The emission count was determined by summing the background subtracted temporal profile of the measured response.
2.3 Non-linear least squares fitting
Non-linear least squares fitting was employed using the Levenberg-Marquadt minimization algorithm in MatLab (Mathworks, Massachusetts—lsqnonlin function). The model decay function was
Here, an and τn are the pre-exponential amplitudes and the time constants of the two decay components respectively. When fitting, the model decay function was convolved with the IRF prior to fitting to the data. The value for χ2 was then determined using the following formula:
We note that the weighting of the data was calculated from the decay model (Pearson’s χ2).
2.4 Fitting the temporal shift between two IRFs
Fitting of the temporal shift between two IRFs was performed by minimizing the weighted mean square difference between the two curves. The weighting used assumed a Poisson noise distribution. Temporal shifts of less than one bin width were introduced by linear interpolation of the original IRF.
To estimate the error of the procedure above, an artificial IRF was generated using a sum of four Gaussians with parameters chosen to mimic the shape of a typical IRF. Two curves were generated from the artificial IRF by adding a background and adjusting the amplitude to match those of the two IRFs under examination. Poisson noise was added to the two curves and the temporal shift between them was fitted. This was then repeated for many different pairs of noisy curves. The mean temporal shift tends toward zero with increasing repeats and the standard deviation of the resulting distribution was used for our estimate of the error on the measured shift.
2.5 Skin sample
The skin sample was obtained from a patient attending the Department of Dermatology at Imperial College Healthcare NHS Trust. The use of tissue was ethically approved by the regional ethics committee and patients gave written informed consent to participate. The clinical diagnosis was confirmed histologically after imaging.
The freshly excised lesion was rinsed with Hanks balanced salt solution (HBSS) buffer without phenol red, calcium or magnesium (Gibco®, Invitrogen, CA, USA) and placed on damp gauze in an inverted glass bottomed petri dish containing a 170 µm cover slip (MatTek®, MA,USA). This was attached to a metallic ring using adhesive and magnetically coupled around the oil immersion microscope objective using Immersol 518F (Carl Zeiss, Germany) at the interface.
3. Results and discussion
3.1 Spectral response of the gold nanorods
Figure 2 shows the extinction spectrum of the gold nanorods, together with the emission spectra measured using the spectrally resolved detector shown in Fig. 1(b). We note that the extinction spectrum is the sum of scattering and absorption. The extinction spectrum shows two peaks, one at 520 nm, corresponding to the transverse surface plasmon resonance [25], and one at 825 nm corresponding to the longitudinal plasmon resonance. The width of these bands is assigned to the interplay of the inhomogeneous distribution of the particles shapes [34,35] and the plasmon-plasmon coupling, which may occur at the high nanorod concentrations used in these samples [38]. It can be seen that the emission spectra are very similar for the different excitation wavelengths used and have a peak at ~520 nm, which is coincidentally similar to the gold luminescence peak [24]. The broad range of the luminescence extends across the spectral detection range of our multispectral TCSPC detector and we note that although it doesn’t extend beyond ~640 nm, the emission spectrum of nanorods can be tuned by altering their aspect ratio [25]. To collect decay profiles from the lifetime standards presented in [23] for use with the reference reconvolution technique would require approximately four separate dyes in order to achieve the same spectral coverage.
Fig. 2 (a) Extinction and (b) emission spectra of the gold nanorods. The extinction spectrum was acquired using the nanorods embedded into a thin film, rather than in solution and therefore the extinction coefficient is in arbitrary units.
3.2 Comparing the gold nanorod IRF to a second harmonic IRF
Figure 3(a) shows the temporal response of the short wavelength channel of the multispectral TCSPC detector to the SHG signal from urea crystals, as well as to the nanorods excited under identical conditions. It can be seen that the two curves are very similar, indicating that the very short luminescence of the gold nanorods provides a very good approximation to a true delta function impulse response for the purposes of determining the IRF. The two curves shown in Fig. 3(a) were found to be nearly identical (χ2 = 1.3) to within the Poisson noise on the measurements, except for a 6.1 ps temporal shift. We attribute this shift to the color shift of the PMT detector employed, as the SHG signal has a narrow spectrum centred at 420 nm, while the nanorod luminescence has a broad emission spectrum across a large portion of the detection band (~370-420 nm). Multiphoton excited luminescence from gold nanorods has previously been shown to have no detectable delay following excitation in measurements carried out with a sub 200 fs temporal resolution [31]. There have been some reports in the literature of decay components with very long luminescence decay times from gold nanorods (0.12 – 2.0 ns) [29,32], however, for the nanorod samples used here, we did not observe these longer components, as can be seen in Fig. 3(a).
Fig. 3 (a) Comparison of the temporal response of luminescence from the gold nanorods excited at 840 nm to that of the SHG signal from urea crystals in the short wavelength channel (360-425 nm) of the multispectral TCSPC detector. (b) the wavelength dependence of the temporal shift (the color shift) of the SHG signal relative to the nanorod luminescence in the short wavelength channel. The error bars are estimates obtained from numerical simulations, see section 2.4.
Figure 3(b) shows the temporal shift of the SHG signal from urea crystals relative to the gold luminescence as a function of excitation wavelength. We note that SHG signals obtained at different excitation wavelengths cannot be directly compared to each other since the timing of the TCSPC stop trigger pulse can be sensitive to the power level and repetition rate of the excitation source, both of which are dependent on wavelength. Therefore, the temporal offset of the SHG signal at each wavelength was measured relative to the luminescence from gold nanorods at the same excitation wavelength. Since the emission spectrum of the nanorods is constant with respect to excitation wavelength, as shown in Fig. 2(b), any color shift effect caused by the detector will not alter the nanorod IRF when the excitation wavelength is changed. Therefore, the temporal offset of the SHG IRF relative to the nanorod IRF, as shown in Fig. 3(b), represents the color shift of this detector at a specific SHG wavelength compared to broadband illumination from gold nanorod luminescence.
As the ultrafast nanorod luminescence is spectrally more similar to the emission spectrum of a typical fluorescent sample than to the narrow band SHG signal (i.e. both the gold nanorods IRF and the sample are acquired with broadband illumination of the PMT), it greatly reduces the effect of the color shift as well as the effects of any wavelength dependent change in the shape of the IRF. Gold nanorod luminescence therefore provides a more realistic direct measurement of the IRF in the actual detection band used, which was not previously possible. In turn, this facilitates the more reliable and precise fitting of exponential decays required for measurements that are reproducible and consistent between different instruments.
3.3 Response of the gold nanorods as a function of excitation power
The dependence of gold nanoparticle luminescence signal on excitation beam intensity has been measured previously by many authors e.g., [28,36,37,45–48]. and shown to depend strongly on particle shape and size. We have investigated the power dependence of the luminescence of our nanorod samples under multiphoton excitation using a Leica TCS FLIM SP5 that allows a convenient monitoring of the excitation power during the experiment. The measurements were performed in order of increasing excitation power to avoid photodamage at high powers. The emission wavelength band was 400–500 nm, and our results are shown in Fig. 4 . Initially, the signal follows a ~2.61 power law dependence, indicating that it is a combination of two and three photon processes, in agreement with previous literature [46]. However, at higher excitation powers, the curve deviates from this relationship which may be due to a morphological change induced by the high excitation power, e.g. due to annealing [49–51].
Fig. 4 Plot of the gold nanorod luminescence count rate as a function of excitation intensity at the sample plane. The fit line was calculated using only the first four data points.
It should be noted that a morphological change of the gold nanoparticles may result in a blue shift of the peak of the emission spectrum by several tens of nanometers [51]. However, this would not affect our experiments since we only require the luminescence to provide an ultrashort signal (<1 ps) and we have confirmed that if we do irradiate with sufficient power to achieve annealing, the temporal FWHM of the luminescence signal is unchanged. In practice, however, we do not use such high powers to excite the nanorods for the measurement of an IRF. As shown in Fig. 2(b), no change in the peak emission wavelength is observed in the same field of view of a nanorod ensemble for consecutive emission spectra acquired from a previously unirradiated region, suggesting that no annealing is taking place. Typically we use an excitation beam power of 15 MW cm−2 at the sample—well below the point at which any deviation is observed in Fig. 4. Furthermore, we note that the 800 nm excitation wavelength used for this measurement would realize 3-photon absorption of a 267 nm transition or 2-photon absorption of a 400 nm transition. Varnavski et al. [31] showed that the emission of gold nanorods following single photon excitation at 267 nm and at 400 nm has a decay time of less than the temporal resolution of their instrument (50 fs).
3.4 Use of gold nanorods to acquire a multi spectral IRF
As discussed in the introduction, the inherently narrow emission spectrum of SHG means that it cannot be used to measure the IRF in every emission channel without modifying the emission filters or the detector. Figure 5 shows the temporal response of the system measured using the nanorod luminescence in the four spectral channels for a wide range of excitation wavelengths. It can be seen that the broad band, ultra-short emission of gold luminescence provides a suitable source for an IRF in each emission channel. Each four channel IRF was collected in a single acquisition (i.e. one acquisition per excitation wavelength) in a very reasonable time period of <5 seconds per excitation wavelength. The slight differences between the responses in the different spectral channels arise from the slightly different optical and electronic paths between the channels as well as differences between the individual PMTs and therefore these curves do not provide quantitative information about the color shifts between the detection channels. Since the emission spectrum of the nanorods does not change with excitation wavelength, we do not expect to observe a color shift in any individual channel when changing the excitation wavelength. However, we do expect to observe a temporal shift which is constant across all four channels due to the dependence of the TCSPC timing on excitation wavelength. This was confirmed, to within the error, by comparing the temporal shift in the IRFs of each channel collected at different excitation wavelengths (e.g. the temporal shift between the IRFs measured in the blue channel at λex = 760 nm and 840 nm were the same as the shift between the IRFs measured in the red channel at the same excitation wavelengths).
Fig. 5 Showing the temporal response of the system measured using the nanorod sample over a wide range of excitation wavelengths measured with the 4-channel detector.
The three sets of IRFs were acquired consecutively from the same field of view within the gold nanorod specimen. The relative amplitudes of the IRFs in each spectral channel are constant between the three acquisitions, indicating that the spectrum of the luminescence is constant and thus indicating that the nanorods have not changed morphology due to exposure to high excitation power.
3.5 Fitting multispectral FLIM data
Figure 6 shows FLIM data acquired using the 4-channel detection system on the DermaInspect® instrument from a pigmented nodular Basal Cell Carcinoma (nBCC) excised from the chest. The photon count in an individual pixel is not sufficient to fit a complex exponential decay, therefore, the fluorescence lifetime image shown in Fig. 6(b) was produced by fitting a single exponential decay at each pixel. An IRF in each spectral channel was acquired using a sample of the gold nanorods in each spectral channel simultaneously.
Fig. 6 Showing FLIM data acquired from ex vivo tissue containing a nBCC excised from the chest. The excitation wavelength was 760 nm and the emission channel for the images shown in (a) & (b) was 425-515 nm. (a) shows the total intensity image with a field of view of 180 × 180 µm2. (b) shows a FLIM image of the single exponential lifetime map following fitting of the data using a nanorod IRF, where the lifetime image has been merged with the intensity image. The scale bar in (a) and (b) is 40 µm. Data from the cell outlined in red in (a) was spatially binned and fitted in three channels (we did not observe a strong fluorescence signal in the fourth channel). The raw data, IRFs and reduced χ2 values are shown in (c-e).
In order to obtain a greater photon count to fit a complex exponential, an individual cell was selected for analysis by manually selecting the region of interest, spatially binning all of the data within that region to obtain one fluorescence decay profile for each emission channel, and then fitting a double exponential decay curve to that data. The temporal shift of the IRF relative to the data was not adjusted during this fitting and was fixed to zero for all three spectral channels. The spectrally resolved fluorescence decay profiles from this cell have a very short lifetime component, which may be due to the presence of melanin, and are presented in Fig. 6(c)–(e). We note that the fit residuals retain some structure, indicating that the decay is more complex than a double exponential, which is expected as both melanin and NADH have at least two decay components. However, fitting a more complex decay model is not warranted given the number of photons collected. For this sample, the fluorescence signal in the longest wavelength channel was very low and so is not shown here.
4. Conclusion
In time-resolved fluorescence decay measurements, one must carefully account for the instrument response function of the detector in order to accurately fit complex decays. Currently, for multiphoton FLIM, the IRF may be acquired at half the excitation wavelength using a hyper-Rayleigh or SHG scattering signal. However, due to the color dependent shift or broadening of the temporal response of the detector, this provides a suboptimal measurement. These techniques are also inconvenient to implement for multispectral measurements.
To address these points, the most practically effective approaches to account for the IRF are to measure the response of the system to a very short lifetime dye and approximate this to the impulse response or to measure the response to a reference fluorophore with a known single exponential decay and employ the reference reconvolution technique. The former approach does not allow short fluorescence decays such as melanin or NAD(P)H to be fitted accurately. The latter approach requires the use of a minimum of two reference fluorophores in order to calibrate all four spectral channels in our multispectral FLIM system, which is not convenient when performing clinical measurements. As an alternative to reference fluorophores, we have presented the use of gold nanorods as an appropriate sample for obtaining broadband IRF measurements for multispectral multiphoton fluorescence lifetime imaging. Their ultrafast and broadband luminescence means that they provide a near-perfect approximation to a delta function impulse, permitting accurate determination of the IRF for fluorescence decay measurements. We have experimentally demonstrated that gold nanorods provide measurements of the IRF across a broad spectral range and therefore can be used to determine the IRF of a multispectral FLIM detector. If the detection or excitation bands of the detector are different to the one described here, the spectral range of the gold nanorod excitation and emission may be modified by altering the distribution of nanorod aspect ratios. Furthermore, the gold nanorods provide a physically robust and photostable specimen that can be used repeatedly in a clinical setting. If the detector employed exhibits a color shift in its response, then the broadband luminescence of the gold nanorods inherently provides a measurement of the IRF over the same detection band that is used to collect the fluorescence signal from a sample, thus minimizing the effect of the color-shift on the lifetime measurement. In conclusion, we have presented the use of gold nanorod luminescence for the accurate acquisition of an IRF in multiphoton FLIM that facilitates the accurate interpretation of data with potential applications in, for example, but not limited to, skin lesion classification in the clinical setting.
Acknowledgments
The authors gratefully acknowledge funding from the European Commission (SKINSPECTION, FP7-HEALTH-2007-A, 201577 and PHOTONICS-4-LIFE, FP7-ICT-2007-2, 224014). SW acknowledges a Ph.D. studentship from the Institute of Chemical Biology EPSRC funded Doctoral Training Centre. The assistance of the Human Biomaterials Resource Centre of Imperial College Healthcare NHS Trust in obtaining tissue for this study is gratefully acknowledged. PF acknowledges a Royal Society Wolfson Research Merit Award. The authors gratefully acknowledge the technical expertise of Marcel Kellner-Höfer and Reiner Bückle.
References and links
1. W. Denk, J. H. Strickler, and W. W. Webb, “Two-photon laser scanning fluorescence microscopy,” Science 248(4951), 73–76 (1990). [CrossRef] [PubMed]
2. K. Koenig and I. Riemann, “High-resolution multiphoton tomography of human skin with subcellular spatial resolution and picosecond time resolution,” J. Biomed. Opt. 8(3), 432–439 (2003). [CrossRef] [PubMed]
3. B. R. Masters, P. T. C. So, and E. Gratton, “Multiphoton excitation fluorescence microscopy and spectroscopy of in vivo human skin,” Biophys. J. 72(6), 2405–2412 (1997). [CrossRef] [PubMed]
4. W. Becker, A. Bergmann, E. Haustein, Z. Petrasek, P. Schwille, C. Biskup, T. Anhut, I. Riemann, and K. Koenig, “Fluorescence lifetime images and correlation spectra obtained by multi-dimensional TCSPC,” in Multiphoton Microscopy in the Biomedical Sciences V, A. Periasamy, and P. T. C. So, eds. (SPIE, Bellingham, 2005), pp. 144–151.
5. A. Pradhan, P. Pal, G. Durocher, L. Villeneuve, A. Balassy, F. Babai, L. Gaboury, and L. Blanchard, “Steady state and time-resolved fluorescence properties of metastatic and non-metastatic malignant cells from different species,” J. Photochem. Photobiol. B 31(3), 101–112 (1995). [CrossRef] [PubMed]
6. M. C. Skala, K. M. Riching, D. K. Bird, A. Gendron-Fitzpatrick, J. Eickhoff, K. W. Eliceiri, P. J. Keely, and N. Ramanujam, “In vivo multiphoton fluorescence lifetime imaging of protein-bound and free nicotinamide adenine dinucleotide in normal and precancerous epithelia,” J. Biomed. Opt. 12(2), 024014 (2007). [CrossRef] [PubMed]
7. K. Teuchner, J. Ehlert, W. Freyer, D. Leupold, P. Altmeyer, M. Stücker, and K. Hoffmann, “Fluorescence Studies of melanin by stepwise two-photon femtosecond laser excitation,” J. Fluoresc. 10(3), 275–282 (2000). [CrossRef]
8. H. D. Vishwasrao, A. A. Heikal, K. A. Kasischke, and W. W. Webb, “Conformational dependence of intracellular NADH on metabolic state revealed by associated fluorescence anisotropy,” J. Biol. Chem. 280(26), 25119–25126 (2005). [CrossRef] [PubMed]
9. Y. Wu, W. Zheng, and J. Y. Qu, “Sensing cell metabolism by time-resolved autofluorescence,” Opt. Lett. 31(21), 3122–3124 (2006). [CrossRef] [PubMed]
10. J. A. Palero, H. S. de Bruijn, A. van der Ploeg-van den Heuvel, H. J. C. M. Sterenborg, and H. C. Gerritsen, “In vivo nonlinear spectral imaging in mouse skin,” Opt. Express 14(10), 4395–4402 (2006). [CrossRef] [PubMed]
11. D. K. Bird, K. W. Eliceiri, C. H. Fan, and J. G. White, “Simultaneous two-photon spectral and lifetime fluorescence microscopy,” Appl. Opt. 43(27), 5173–5182 (2004). [CrossRef] [PubMed]
12. A. Habenicht, J. Hjelm, E. Mukhtar, F. Bergstrom, and L. B. A. Johansson, “Two-photon excitation and time-resolved fluorescence: 1. The proper response function for analysing single-photon counting experiments,” Chem. Phys. Lett. 354(5-6), 367–375 (2002). [CrossRef]
13. W. Becker, “Recording the instrument response function of a multiphoton FLIM sytem,” Becker & Hickl GmbH Application Note irf-mp-04.doc (2008), http://www.becker-hickl.de/pdf/irf-mp04.pdf.
14. R. Cicchi, L. Sacconi, A. Jasaitis, R. P. O’Connor, D. Massi, S. Sestini, V. De Giorgi, T. Lotti, and F. S. Pavone, “Multidimensional custom-made non-linear microscope: from ex-vivo to in-vivo imaging,” Appl. Phys. B 92(3), 359–365 (2008). [CrossRef]
15. E. Dimitrow, I. Riemann, A. Ehlers, M. J. Koehler, J. Norgauer, P. Elsner, K. König, and M. Kaatz, “Spectral fluorescence lifetime detection and selective melanin imaging by multiphoton laser tomography for melanoma diagnosis,” Exp. Dermatol. 18(6), 509–515 (2009). [CrossRef] [PubMed]
16. J. R. Lakowicz, Principles of Fluorescence Spectroscopy (Kluwer Academic/Plenum, New York, 1999).
17. D. Bebelaar, “Time response of various types of photomultipliers and its wavelength dependence in time-correlated single-photon counting with an ultimate resolution of 47 ps FWHM,” Rev. Sci. Instrum. 57(6), 1116–1125 (1986). [CrossRef]
18. R. Krahl, A. Bülter, and F. Koberling, “Performance of the Micro Photon Devices PDM 50CT SPAD detector with PicoQuant TCSPC systems” Technical Note (PicoQuant GmbH, 2005).
19. R. Luchowski, M. Szabelski, P. Sarkar, E. Apicella, K. Midde, S. Raut, J. Borejdo, Z. Gryczynski, and I. Gryczynski, “Fluorescence instrument response standards in two-photon time-resolved spectroscopy,” Appl. Spectrosc. 64(8), 918–922 (2010). [CrossRef] [PubMed]
20. K. Teuchner, W. Freyer, D. Leupold, A. Volkmer, D. J. S. Birch, P. Altmeyer, M. Stücker, and K. Hoffmann, “Femtosecond two-photon excited fluorescence of melanin,” Photochem. Photobiol. 70(2), 146–151 (1999). [PubMed]
21. M. Wakita, G. Nishimura, and M. Tamura, “Some characteristics of the fluorescence lifetime of reduced pyridine nucleotides in isolated mitochondria, isolated hepatocytes, and perfused rat liver in situ,” J. Biochem. 118(6), 1151–1160 (1995). [PubMed]
22. M. Zuker, A. G. Szabo, L. Bramall, D. T. Krajcarski, and B. Selinger, “Delta-function convolution method (DFCM) for fluorescence decay experiments,” Rev. Sci. Instrum. 56(1), 14–22 (1985). [CrossRef]
23. N. Boens, W. W. Qin, N. Basarić, J. Hofkens, M. Ameloot, J. Pouget, J. P. Lefèvre, B. Valeur, E. Gratton, M. vandeVen, N. D. Silva Jr, Y. Engelborghs, K. Willaert, A. Sillen, G. Rumbles, D. Phillips, A. J. Visser, A. van Hoek, J. R. Lakowicz, H. Malak, I. Gryczynski, A. G. Szabo, D. T. Krajcarski, N. Tamai, and A. Miura, “Fluorescence lifetime standards for time and frequency domain fluorescence spectroscopy,” Anal. Chem. 79(5), 2137–2149 (2007). [CrossRef] [PubMed]
24. A. Mooradian, “Photoluminescence of Metals,” Phys. Rev. Lett. 22(5), 185–187 (1969). [CrossRef]
25. M. B. Mohamed, V. Volkov, S. Link, and M. A. El-Sayed, “The 'lightning' gold nanorods: fluorescence enhancement of over a million compared to the gold metal,” Chem. Phys. Lett. 317(6), 517–523 (2000). [CrossRef]
26. E. Dulkeith, T. Niedereichholz, T. A. Klar, J. Feldmann, G. von Plessen, D. I. Gittins, K. S. Mayya, and F. Caruso, “Plasmon emission in photoexcited gold nanoparticles,” Phys. Rev. B 70(20), 205424 (2004). [CrossRef]
27. C. Sönnichsen, T. Franzl, T. Wilk, G. von Plessen, J. Feldmann, O. Wilson, and P. Mulvaney, “Drastic reduction of plasmon damping in gold nanorods,” Phys. Rev. Lett. 88(7), 077402 (2002). [CrossRef] [PubMed]
28. M. R. Beversluis, A. Bouhelier, and L. Novotny, “Continuum generation from single gold nanostructures through near-field mediated intraband transitions,” Phys. Rev. B 68(11), 115433 (2003). [CrossRef]
29. S. Park, M. Pelton, M. Liu, P. Guyot-Sionnest, and N. F. Scherer, “Ultrafast resonant dynamics of surface plasmons in gold nanorods,” J. Phys. Chem. C 111(1), 116–123 (2007). [CrossRef]
30. O. Varnavski, R. G. Ispasoiu, L. Balogh, D. Tomalia, and T. Goodson, “Ultrafast time-resolved photoluminescence from novel metal-dendrimer nanocomposites,” J. Chem. Phys. 114(5), 1962–1965 (2001). [CrossRef]
31. O. P. Varnavski, M. B. Mohamed, M. A. El-Sayed, and T. Goodson, “Relative enhancement of ultrafast emission in gold nanorods,” J. Phys. Chem. B 107(14), 3101–3104 (2003). [CrossRef]
32. K. Imura, T. Nagahara, and H. Okamoto, “Near-field two-photon-induced photoluminescence from single gold nanorods and imaging of plasmon modes,” J. Phys. Chem. B 109(27), 13214–13220 (2005). [CrossRef] [PubMed]
33. G. T. Boyd, Z. H. Yu, and Y. R. Shen, “Photoinduced luminescence from the noble metals and its enhancement on roughened surfaces,” Phys. Rev. B Condens. Matter 33(12), 7923–7936 (1986). [CrossRef] [PubMed]
34. S. Eustis and M. A. el-Sayed, “Why gold nanoparticles are more precious than pretty gold: noble metal surface plasmon resonance and its enhancement of the radiative and nonradiative properties of nanocrystals of different shapes,” Chem. Soc. Rev. 35(3), 209–217 (2006). [CrossRef] [PubMed]
35. J. Tao, Y.-H. Lu, R.-S. Zheng, K.-Q. Lin, Z.-G. Xie, Z.-F. Luo, S.-L. Li, P. Wang, and H. Ming, “Effect of aspect ratio distribution on localized surface plasmon resonance extinction spectrum of gold nanorods,” Chin. Phys. Lett. 25(12), 4459–4462 (2008). [CrossRef]
36. H. F. Wang, T. B. Huff, D. A. Zweifel, W. He, P. S. Low, A. Wei, and J. X. Cheng, “In vitro and in vivo two-photon luminescence imaging of single gold nanorods,” Proc. Natl. Acad. Sci. U.S.A. 102(44), 15752–15756 (2005). [CrossRef] [PubMed]
37. Y. A. Zhang, J. Yu, D. J. S. Birch, and Y. Chen, “Gold nanorods for fluorescence lifetime imaging in biology,” J. Biomed. Opt. 15(2), 020504 (2010). [CrossRef] [PubMed]
38. P. Matteini, F. Ratto, F. Rossi, S. Centi, L. Dei, and R. Pini, “Chitosan films doped with gold nanorods as laser-activatable hybrid bioadhesives,” Adv. Mater. (Deerfield Beach Fla.) 22(38), 4313–4316 (2010). [CrossRef] [PubMed]
39. J. Pérez-Juste, B. Rodriguez-Gonzalez, P. Mulvaney, and L. M. Liz-Marzan, “Optical control and patterning of gold-nanorod-poly(vinyl alcohol) nanocomposite films,” Adv. Funct. Mater. 15(7), 1065–1071 (2005). [CrossRef]
40. P. Zijlstra, J. W. M. Chon, and M. Gu, “Five-dimensional optical recording mediated by surface plasmons in gold nanorods,” Nature 459(7245), 410–413 (2009). [CrossRef] [PubMed]
41. F. Ratto, P. Matteini, F. Rossi, and R. Pini, “Size and shape control in the overgrowth of gold nanorods,” J. Nanopart. Res. 12(6), 2029–2036 (2010). [CrossRef]
42. L. F. Gou and C. J. Murphy, “Fine-tuning the shape of gold nanorods,” Chem. Mater. 17(14), 3668–3672 (2005). [CrossRef]
43. X. D. Xu and M. B. Cortie, “Shape change and color gamut in gold nanorods, dumbbells, and dog bones,” Adv. Funct. Mater. 16(16), 2170–2176 (2006). [CrossRef]
44. C. B. Talbot, R. Patalay, I. H. Munro, H. G. Breunig, K. Konig, Y. Alexandrov, S. Warren, A. Chu, G. W. Stamp, M. A. A. Neil, P. M. W. French, and C. W. Dunsby, “A multispectral FLIM microscope for in-vivo imaging of skin cancer,” P. Ammasi, K. Karsten, and T. C. S. Peter, eds. (SPIE, 2011), p. 79032B.
45. P. Biagioni, M. Celebrano, M. Savoini, G. Grancini, D. Brida, S. Mátéfi-Tempfli, M. Mátéfi-Tempfli, L. Duò, B. Hecht, G. Cerullo, and M. Finazzi, “Dependence of the two-photon photoluminescence yield of gold nanostructures on the laser pulse duration,” Phys. Rev. B 80(4), 045411 (2009). [CrossRef]
46. M. Eichelbaum, B. E. Schmidt, H. Ibrahim, and K. Rademann, “Three-photon-induced luminescence of gold nanoparticles embedded in and located on the surface of glassy nanolayers,” Nanotechnology 18(35), 355702 (2007). [CrossRef]
47. R. A. Farrer, F. L. Butterfield, V. W. Chen, and J. T. Fourkas, “Highly efficient multiphoton-absorption-induced luminescence from gold nanoparticles,” Nano Lett. 5(6), 1139–1142 (2005). [CrossRef] [PubMed]
48. L. Tong, C. M. Cobley, J. Chen, Y. Xia, and J.-X. Cheng, “Bright three-photon luminescence from gold/silver alloyed nanostructures for bioimaging with negligible photothermal toxicity,” Angew. Chem. Int. Ed. Engl. 49(20), 3485–3488 (2010). [CrossRef] [PubMed]
49. S. Link, C. Burda, B. Nikoobakht, and M. A. El-Sayed, “Laser-induced shape changes of colloidal gold nanorods using femtosecond and nanosecond laser pulses,” J. Phys. Chem. B 104(26), 6152–6163 (2000). [CrossRef]
50. S. Link, Z. L. Wang, and M. A. El-Sayed, “How does a gold nanorod melt?” J. Phys. Chem. B 104(33), 7867–7870 (2000). [CrossRef]
51. A. Bouhelier, R. Bachelot, G. Lerondel, S. Kostcheev, P. Royer, and G. P. Wiederrecht, “Surface plasmon characteristics of tunable photoluminescence in single gold nanorods,” Phys. Rev. Lett. 95(26), 267405 (2005). [CrossRef] [PubMed]
