Many naturally occurring substances are autofluorescent when excited with UV or visible wavelengths. Autofluorescence emission typically spans the visible spectrum with a lifetime measured in nanoseconds. One of the earliest reports of using a probe fluorescence lifetime to discriminate against nonspecific background autofluorescence was made by Thaer and Sernetz in 1973.1 Since then, a number of microscopes with the ability to resolve different fluorophores on the basis of have been reported. 2, 3, 4, 5 Instruments that operate in the time domain to resolve fluorophores that differ in by a large degree (ns versus ) have the advantage of simplicity and lower cost compared to microscopes required to resolve fluorophores on the basis of a few nanoseconds’ difference. The time-gated luminescence (TGL) microscope described here operates within the time domain to capture long-lived (greater than ) emission after autofluorescence has decayed. Figure 1 illustrates the basic concept of TGL; with the detector off, the TGL cycle begins with a short, powerful excitation pulse that raises the target luminophore into its excited state. On termination of the excitation pulse, nonspecific fluorescence decays rapidly while target luminescence persists for orders of magnitude longer. After a resolving period (gate delay), the detector is gated on (acquisition period) to capture luminescent emission in the absence of autofluorescence.
TGL microscopes employ a pulsed excitation scheme at a wavelength suited to the target luminophore. Platinum and palladium porphyrin based luminophores can be excited at either 390 or ; a number of solid-state or semiconductor excitation sources are suitable for this role. Unlike the former compounds, lanthanide chelates are not oxygen sensitive and can provide longer lifetimes (0.5 to ). They are normally employed with the ions bound to a sensitizer molecule to boost the absorbance cross-section. Typically they require excitation in the UV region of the spectrum ( : ; : ), with the upper useful limit being about for terpyridine-based europium chelates ( effective at ). Some europium chelates in novel configurations can be excited at longer wavelengths (365 to ), although they were not used for this work due to other important limitations. 6, 7, 8, 9 The luminescence lifetime of europium chelates is typically around 300 to in aqueous environments and follows single exponential decay kinetics.
Microscopes designed for use with lanthanide chelates usually employ pulsed UV sources such as Xe flashlamps, or nitrogen-laser or chopper-interrupted Hg arc lamps. 10, 11, 12, 13, 14, 15 As a consequence of their low duty cycle, phosphors emit relatively weakly compared with most fluorophores and therefore require sensitive detectors. All previously reported TGL microscopes have required multiple excitation detection cycles to deliver an image of acceptable contrast and quality. The detector integrates photons over many TGL cycles, and it is necessary to shield the sensor from light during the excitation cycle. Microchannel-plate image intensifiers that employ electronic gating are used to satisfy this requirement, whereas conventional CCD cameras require an external shutter mechanism. Regardless of the technique used, multiple excitation cycles have been necessary, requiring either an expensive gated image intensifier, 11, 14, 15, 16, 17, 18, 19, 20, 21 a vibration-prone mechanical beam interruptor (chopper),12, 22 or a high-insertion-loss ferro-electric LCD shutter 14, 20 to control the light reaching the detector.
We previously reported the design of a UV LED-excited TGL microscope for use with europium fluorophores.23 The recent availability of electron-multiplying CCD (EM-CCD) cameras prompted us to consider their application in TGL microscopy. EM-CCD cameras are the solid-state equivalent of image-intensified CCD cameras, albeit with lower gain, which is compensated to some extent by a threefold improvement in quantum sensitivity.
While EM-CCD cameras offer high sensitivity, they still require an external shutter mechanism if multiple TGL cycles are to be employed for each acquisition. Alternatively, if a single exposure cycle is sufficient, as with the system described here, the shutter can be eliminated.
Due to the small size of the LED, the device was mounted within an Olympus U-MWU2 filter cube as shown in Fig. 2 . The UV LED (NCSU033A, Nichia Corp., Japan) used for this work was an improved device rated at at , about double the output power of the previous version (NCCU033). The LED was surface-mounted to a 25-mm-diameter single-sided printed circuit board (PCB) that replaced the excitation filter in the cube. In the confined space of the cube, it was not possible to achieve Koehler illumination and a diffuser (frosted glass slide) was mounted on the front face of the LED to homogenize the beam. To the eye, the excitation region appeared uniform in intensity, although scatter from the diffuser was estimated to reduce output power by about 15%. Power was supplied to the LED via flexible power leads that entered the filter housing at its central axis to permit filter cubes on either side of the UV LED to be rotated into view. The filter housing was thus limited to rotation from the UV LED axis due to the length of the power leads.
Earlier we reported the existence of low-intensity self-excited visible luminescence from -based LEDs that persists for some time following switch-off and that can present a problem when the devices are used in pulse fluorometry applications.24, 25 To suppress this component, a short-pass filter (Hoya U-360, Edmund Optics, Singapore) was included in the excitation beam path that was situated about from the LED face, as shown in Fig. 2. A fused silica lens , (LA4647-UV, Thorlabs, Newton, New Jersey) was mounted approximately from the LED face to collimate the excitation radiation. The filtered excitation beam was then directed into the microscope objective via the DM400 dichroic mirror. This mirror strongly reflects wavelengths below while transmitting visible light better than 90%. The increased optical clarity of this arrangement helped maximize fluorescence and excitation efficiency of the instrument.
An Olympus BX51 fluorescence microscope was used for this work, and images were acquired without the benefit of spectral filtering. The UV-LED was supplied from a programmable voltage source so that it could be driven at two different power levels. The LED current was monitored by measuring the peak voltage across the 5-ohm LED load resistor with an oscilloscope; in low-power mode, the current was at a supply voltage of , and in high-power mode, it was 1.44 amps at . The LED was always operated in pulsed mode, and it was convenient to switch to low-power mode to limit photobleaching when higher-duty cycles or long observation periods were employed.
LED Output Power
A Coherent FieldMax™ -TO laser power meter fitted with a model PS10Q detector head was used for power measurements. The output face of the U-MWU2 filter cube was fixed approximately from the PS10Q sensor for power measurements. The pulse profile for both the TGL and epifluorescence modes is shown in Fig. 3 . The rising edge of the trigger pulse would shift from the start to the end of the UV LED pulse when it was switched from prompt (epifluorescence) mode to TGL mode. The duration of the LED pulse was with a 51.6-ms resting period between each pulse, corresponding to a frequency of and a duty cycle of 1:63. The observed average power was with a calculated peak power of . In idle mode , the observed average power was with a calculated peak power of . The LED (still fitted with the diffuser) was then removed from the filter cube and placed from the sensor. In TGL mode, the average power reading was with a peak power of , or 3.05-fold higher than when mounted in the cube. In idle mode, the average power was with a 17.4-mW peak power, corresponding to 2.6 times thepower level when mounted in the cube.
By comparison, our previously reported UV-LED filter assembly, which lacked both the diffuser and the Hoya 360 filter, delivered an average power of with a 29.72-mW peak (LED in filter cube) when measured with the FieldMax™ power meter.23
An iXon DV885 EM-CCD was fitted to the microscope using a standard C-mount lens adaptor. The DV885 camera specifications include: Texas Instruments Impactron frame transfer CCD sensor, pixels, EM gain 2000, quantum efficiency of 65% at , 14-bit digitized output, 24 full frames per second, and external trigger mode support. An embedded microcontroller was used to control the camera, which was operated in “fast external trigger” mode so the instrument could be switched instantly between conventional “prompt” fluorescence and TGL modes. To make the system more versatile, a microcontroller was used to control the gate-delay interval, repetition frequency, trigger pulse polarity, LED pulse length, and drive intensity parameters. In fast-trigger mode the sensor and its registers are cleared pending the arrival of the trigger pulse, the rising edge of which initiates frame capture within nanoseconds. A gate delay of was imposed between the termination of the LED pulse and the rising edge of the trigger pulse.
Giardia lamblia cysts (Biotech Frontiers Pty. Ltd., Sydney, Australia) were labeled using the europium chelate BHHST ( heptafluoro- hexanedion- sulfonyl-aminopropyl-ester-N-succinimide-ester-o-terphenyl), the synthesis and use of which has previously been reported.26 The 10,000:1 concentrate isolated from the Sydney water supply used for this work was a kind gift from Dr. Belinda Ferrari and was prepared from backwash water samples using the flocculation method.27 To further increase the autofluorescence background, the UV excitable fluorophore 7-dimethylaminocoumarin-4-acetic acid (DMACA) was added to the water concentrate together with the Giardia cysts.
Results and Discussion
The addition of DMACA resulted in strong background fluorescence that limited visibility of the Giardia cyst situated at the bottom-left of Fig. 4a . This 8-bit image was acquired using a 40˟ objective, a 3-ms exposure with averaging enabled (4 x frames), and EM gain turned off. Figure 4b was acquired after the microscope was switched to TGL mode and EM gain was increased to 1185. The line A-B transects the cyst in both the prompt and TGL capture frames to generate the profile shown in Fig. 5a . For background determination, the second line profile C-D was sampled; pixel values for this trajectory are shown in Fig. 5b. Data points from these two sets were analyzed to determine the effective improvement in the signal-to-noise ratio (SNR), and key values used for this calculation are shown in Table 1 . The SNR figures were based on the average 8-bit intensity value of the cyst referenced to the brightest region of nonspecific fluorescence within the frame. In prompt epifluorescence mode, the cyst emitted weakly in comparison with other regions and a SNR of was obtained. In TGL mode, the cyst was the only object visible and the SNR improved 30-fold to a value of . The relatively large error bars arise from the small sample size of 17 (for the cyst) with pixel values ranging from 104 to 174.
Summary of the input values used to calculate the improvement in SNR for the Giardia cyst shown in Fig. 4. To compare Figs. 4a and 4b, the SNR was determined by sampling identical regions and calculating the ratio of the brightest signal to the brightest (autofluorescent) background. The effective improvement was taken as the ratio of TGLSNR to PROMPTSNR .
Effect of Frame Averaging
Software supplied with the iXon camera provided the option to average successive frames and improve image quality through reduction of random noise components. Averaging significantly improved the SNR by decreasing the background noise level. For example, a background region (sample pixels) within a single frame acquired under TGL conditions had an average value of . When the same frame was averaged over four successive frames, the background dropped to . Increasing the frame count to 8 resulted in a further small improvement in SNR (about 7%).
The improvement in signal strength achieved by frame averaging was determined by monitoring an oval region (sample pixels) on a Giardia cyst present within the frames captured for background measurements. The mean pixel value for the region after a single acquisition was , and this was improved to when four successive frames were averaged (data not shown).
EM Gain and its Effect on SNR
BHHST is a strongly luminescent europium chelate that was conjugated to the anti-Giardia monoclonal antibody G203 for the detection of Giardia lamblia cysts. The iXon camera employs a very sensitive sensor, and it was possible to capture images of well-labeled Giardia cysts even without the assistance of EM gain. Figure 6a was captured in conventional epifluorescent mode and shows an image of a Giardia cyst suspended within a background of fluorescent DMACA. The line profile at the top reports pixel intensity from left to right across the three frames [6a to 6c], and the fluid meniscus and center of the cyst have roughly equal (peak) pixel intensities. Figure 6b was captured in TGL mode with EM gain turned off. Referring again to the line profile, it is apparent that fluorescence from the DMACA was strongly suppressed, and the SNR improved from an initial value of about 1 to 4.6. The crescent at the top of the cyst was an artifact arising from scattered luminescence focused by the meniscus. EM gain (576) was enabled to acquire the image shown in Fig. 6c that had significantly reduced background levels compared with Figs. 6a and 6b. The SNR for this image was improved to around 28 (199/7) by virtue of background suppression and signal strength enhancement delivered by the camera with EM gain enabled.
Our results support the conclusion that substantial improvements in SNR can be achieved in TGL mode without a shutter when EM gain and frame averaging are employed. Increased optical throughput to detector, decreased instrument complexity, and finer control of the gate-delay interval (to maximize detection efficiency) are key benefits arising from the elimination of the shutter.
The solid-state instrument described here implemented a short gate-delay to capture target luminescence at maximal intensity. Good image quality was achieved after a single excitation cycle of when camera EM gain was enabled. While the excitation and exposure portion of a TGL cycle are essentially complete after , the acquisition process must be extended to to allow for the frame readout time. This interval is still faster than the time taken for a motorized stage to ramp up to speed, move to a new location, and stabilize.
An important feature of TGL techniques is the reduction in image complexity that facilitates the use of computer recognition systems to process images for the identification of target organisms based on their morphology. We intend to investigate these techniques for the automated detection of methicillin-resistant Staphylococcus Aureus (MRSA) in sputum samples.
For TGL microscopy, the introduction of inexpensive solid-state LED excitation sources was an exciting development, and the recent availability of EM cameras was equally significant. The cost of implementing TGL microscopy has plummeted while image resolution, SNR, and acquisition rates have improved greatly. With solid-state instrumentation, we believe that TGL microscopy has finally come of age.
We wish to thank the Australian Research Council (ARC) and Olympus Australia for their generous assistance and support under the ARC Linkage Program (LP0775196).