We describe a very simple but elegant approach to two-photon fluorescence signal enhancement by intensity modulation with immediate application in two-photon laser-scanning fluorescence microscopy. This method of enhancement shows potential application in any microscopic technique that result from non-linear photon absorption and plays a pivotal role in live cell imaging.
One of the state of the art technologies of modern day optical imaging, laser-scanning two-photon fluorescence microscopy (Denk et al., 1990) is based on two-photon absorption (TPA) predicted back in 1931 (Goeppert-Mayer, 1931) and experimentally first demonstrated in 1961 (Kaiser & Garrett, 1961) shortly after the advent of lasers. TPA is a third-order non-linear optical phenomenon where two photons are absorbed simultaneously and the probability of absorption varies as the second power of the intensity (I) of incident radiation (Boyd, 1992).
In practice continuous-wave (cw) lasers are not used for non-linear optical microscopy as very high intensity of the incident radiation is required at the sample. This is taken care of by using high-repetition rate lasers producing ultrafast pulses with very high peak power at moderate time-averaged power. Now, the typical time lapse between the pulses of a high-repetition rate laser is nearly 10 ns. On the other hand, in a laser-scanning non-linear optical microscope the images are constructed in pixel by pixel manner and the average laser dwell time per pixel is 10 μs. Therefore, each pixel is illuminated by a bunch of ∼1000 ultra-short pulses and we can consider the illumination to be a quasi continuous one (Denk et al., 2006).
In this paper, we discuss the huge signal enhancement in two-photon fluorescence laser scanning microscopy by intensity modulation of a train of ultrafast laser pulses. Based on our earlier work (De & Goswami, 2009a), we demonstrate how this technique leads to significant fluorescence enhancement with reduced photo damage.
Materials and Methods
In our experiment, we used ∼100 fs pulsed excitation centred on 780 nm from a mode-locked Ti:saph laser (Mira900-F pumped by 532 nm excitation from Verdi5, Coherent, Inc., Portland, OR, U.S.A.) having 76 MHz pulse repetition rate. A neutral density filter wheel was used to control the laser power. We passed the beam through an electro-optic amplitude modulator (4101, driven by a fixed frequency driver, 3363, New Focus, San Jose, CA, U.S.A.) operating at 1 MHz. When a Glan–Taylor prism-polarizer (analyser) is kept immediately after the amplitude modulator, the output signal intensity (Iout) is modulated as:
where the suffix ‘in’ stands for inputs and V's are the voltages with Vπ 19 V when operating at 633 nm. Figure 1 shows the oscilloscope trace of the output of the driver operated at maximum modulation depth. For chopping at 1 kHz, we replaced the electro-optic modulator a mechanical chopper (MC1000A, Thorlabs, Inc., Newton, NJ, U.S.A.). Finally we sent the laser beam to the scan-head of the multiphoton-ready confocal microscope system (FV300 coupled with IX71, Olympus, Inc., Tokyo, Japan) which focuses the beam onto the sample by a high numerical aperture (NA) oil-immersion objective (60×, 1.42 NA). Each of the images comprised 512 × 512 pixels and was acquired by a single scan (with a scan speed of three frames per second) using fluoview software. Images of bovine pulmonary artery endothelial cells (F14781, Molecular Probes, Inc., Eugene, OR, U.S.A.) with nuclei stained by DAPI (4′,6-diamidino-2-phenylindole, blue fluorescent) and α-tubulin by BODIPY FL (4,4-difluoro-4-bora-3a,4a-diaza-s-indacene) goat anti-mouse IgG (immunoglobin-G) (green fluorescent) were acquired by collecting the respective fluorescence using suitable band-pass filters.
Results and discussions
If we blank a cw laser beam with a mechanical chopper having 50% duty cycle (i.e. having 1:1 mark/space ratio) we get pulsed (or chopped) excitation (Fig. 2a) with pulse width equal to the time lapse between the pulses (which is inverse of the pulse repetition rate). Now compare the effects induced by such a pulsed excitation and a cw one having the same time-averaged power (Fig. 1b); for the pulsed excitation the time-averaged power goes down to half of the peak power due to the 50% duty cycle. Let the time-averaged power be I in both cases and the time taken for one complete cycle (i.e. time period) of blanking to be equal to 2T (Fig. 2a). During one cycle the probability of TPA will be:
(the constant directly relates to the TPA cross-section). Similarly while using a cw excitation for the same time window (shown in dashed lines in Fig. 2b) the TPA probability will be:
Thus using a pulsed excitation we have a 2-fold more absorption as compared to the cw excitation when the time-averaged power is the same. As obvious, such enhancement is absent if the process relies on single-photon (i.e. linear dependence on I) absorption; this means the technique is applicable only for non-linear photon absorption. Similarly, it is possible to have a 4-fold more absorption if the absorption is proportional to the third power of the incident radiation, i.e. three-photon absorption (Maiti et al., 1997). Therefore, in general if the absorption varies with the nth power of the incident radiation, we have a 2n−1-fold enhancement. It is interesting to note here that the factor 2 comes due to the 1:1 mark/space ratio of the mechanical chopper; so one can easily play with the amount of enhancement simply by playing with the duty cycle of the mechanical chopper. Also it must be noted that such enhancement could be reflected in all other processes that results from non-linear photon absorption, e.g. TPA microscopy (Ye et al., 2006).
In multiphoton fluorescence laser-scanning microscopy ultrafast lasers, producing train of femtosecond pulses, have been a common choice to circumvent the low multiphoton absorption cross-sections of common fluorophores by making use of the gigantic pulse peak power. If we double the laser power and chop the train of laser pulses, the excitation will comprise bunch of pulses which has the same average power as that of the unchopped one in case of 1:1 blanking (De & Goswami, 2009a). Also, if the blanking frequency is several orders of magnitude slower than the laser repetition rate, the excitation will be almost quasi-continuous and blanking will therefore produce rectangular temporal envelops (each envelope representing the above mentioned bunch of pulses), as shown in Fig. 2. However, in reality this envelope looks much like that shown in Fig. 1 which is due to finite rise and fall times of a chopper.
Now, the 10 ns (which is actually 13 ns in our experiment) inter-pulse time lag matches with the typical excited state life-time (i.e. the time constant for single exponential decay) of most of the fluorophores (Diaspro & Sheppard, 2002; Denk et al., 2006), i.e. decay of excited state population to ground state almost completes within this time window [e.g. the laser dye rhodamine-6G has an excited state lifetime of 4 ns in water (Zimmerman et al., 1982)]. However prolonged exposure to a train of pulses causes photothermal damage to the sample (Sheppard & Kompfner, 1978; Brakenhoff et al., 1996) which results from long time pile up heating effects of the rather transparent (i.e. having negligible linear absorption) solvent contributed by myriads of laser pulses in a pulse train over a long period of time and can be satisfactorily explained by the heat transfer dynamics of the solvent as evident from our previous studies (De & Goswami, 2009a). This happens due to very high instantaneous power of a mode-locked laser pulses compared to the time-averaged power of the pulse train and such solvent induced heating effects are mainly attributed to the higher order non-linear processes (Whinnery, 1974). It is of great importance to note here that this kind of effective removal of photothermal effect persists even at single-photon absorption as it depends only on solvent parameters. However, such effects become important at high power levels (De & Goswami, 2009b), in particular under tight focusing conditions (De & Goswami, in preparation).
We tried to combine the idea of signal enhancement by non-linear photon absorption and laser-scanning fluorescence microscopy. Keeping the 10 μs dwell time in mind, we blanked the excitation at 1 MHz (such that each pixel experience almost 10 cycles of modulation) and compared it with the unblanked excitation having the same average power (Fig. 3). As mechanical choppers can run only up to a few kHz frequencies, we performed our experiment with an electro-optic modulator (EOM). The choice of EOM over an acousto-optic modulator is also due to the greater efficiency of an EOM than an acousto-optic modulator. As shown in Fig. 4, we observed significant fluorescence enhancement when we used blanked excitation over unblanked one, both having the same time-averaged power (the integrated intensity turned out to be more than 2-fold because the modulations depth was less than 100%). It is noteworthy that few commercial laser scanning microscopy (LSM) systems claim to have reduced photothermal etc effects using kHz optical choppers. Under such condition, as the blanking time (few milliseconds) is in between the image acquisition time and the pixel dwell time (Fig. 5), the image is constructed by alternate series of bright and dark pixels (Fig. 6). However, averaging over few such images leads to construction of equivalent images with similar 2-fold signal enhancement which is quite misleading as the pixels experience either quite high or no photo damage as depicted in Fig. 5. Thus, following the logic discussed above, such slow intensity modulation cannot lead to any reduced photo damage as intensity modulation at each pixel is altogether absent.
We blanked 20 mW (time-averaged power) excitation to get modulated excitation at 10 mW and compared it to the unmodulated excitation at 10 mW. These power levels are much below the photo-damage levels as evident from the TPA probability given by (Denk et al., 1990; Diaspro & Sheppard, 2002)
where δ2 is the 2PA cross-section, Pa is the average laser power, τ is the pulse width, f is the pulse repetition rate and NA is the numerical aperture of the focusing objective with h and c having their usual meaning. Also even under the tight-focusing condition the use of oscillator level ∼0.1 nJ pulses (instead of micro-Joule or higher energy pulses from amplified sources) rules out the possibility of pulse-saturation effects which tends to show up as the peak irradiance (i.e. the instantaneous pulse peak power per unit area) reaches GW/cm2 levels (He et al., 2008).
To conclude we have demonstrated that use of blanked radiation provides high signal-to-noise ratio due to the square dependence of the absorption on photon intensity; at the same time blanked excitation with low time-averaged power ensures minimal photothermal damage of the sample during prolonged exposure. This simple trick can be applied for signal enhancement in any other processes that accompany non-linear photon absorption and particularly in live specimen imaging due to its reduced deleterious effect.
AKD thanks CSIR, India for a graduate fellowship. The authors thank DST, India and Wellcome Trust Foundation, UK for funding.