High-Resolution Line-Scanning Two-Photon Microscope

Abstract

A two-photon fluorescence microscope employing line-shaped illumination is presented. This type of excitation is commonly expected to bring about the degradation of axial resolution because of the weaker focusing of the illuminating beam in just one direction. On the basis of a detailed theoretical investigation of the beam shaping performed by cylindrical lenses when inserted in conventional point-scanning systems, we design and implement a microscope set-up making use of readily available optical components. The experimental results show that the proper choice and arrangement of the cylindrical lenses that we have devised is able to preserve the optical-sectioning capability at the video-rate acquisition speed.

1. Introduction

Since the first demonstration of a multiphoton microscope back in 1990 [1], nonlinear fluorescence microscopy has quickly been established as a powerful and reliable technique for noninvasive imaging inside highly scattering media. When compared to confocal fluorescence microscopy, multiphoton fluorescence microscopy allows us to ease, to a large extent, the issues related to phototoxicity, out-of-focus flare, limited imaging depth, and photo-bleaching [2]. Moreover, the wavelengths required for one-photon excitation of most fluorophores lie close to or even inside the ultraviolet range, where laser sources and optical components can be hard to obtain and quite expensive. One of the main drawbacks of a common nonlinear fluorescence microscope is the long acquisition time. As the laser beam is focused down to a tiny spot in order to achieve the high intensity which is necessary for multiphoton excitation, fluorescence is generated from only a very small volume of the sample, which can be assessed as the focal volume. Two-dimensional imaging thus requires a scan of the focal spot in the region of interest, usually by means of galvanometric mirrors, so that an off-line reconstruction of the image is possible. This process may take up to a few seconds, a time range which is too long for the dynamic investigation of several biological processes. Moreover, the low quantum yield makes it commonly necessary to use complex photon-counting detectors.

Many approaches have thus been devised and proposed in order to increase the imaging speed [3]. Optimized scan patterns, such as the spiral one and the Lissajous one, can provide only a limited improvement over the common raster scan. Higher acquisition rates can be obtained either by means of faster scanning devices or by multifocal techniques. The most straightforward solution is line-shaped excitation: the light beam emitted by the laser first passes through suitable cylindrical lenses and is then focused inside the sample with a continuous line of a transversal shape. In such a way the beam has to be scanned along only one transverse direction, and high acquisition rates can be achieved by means of just one conventional scanning galvanometer mirror. Indeed this approach was the first one to be proposed back in 1995 [4], but the poor axial resolution that was recorded soon led to almost complete abandonment. The performance degradation in terms of resolution usually recorded in line-scanning microscopes has been ascribed since then to the fact that the beam is focused in just one direction, thus worsening the optical-sectioning capability while being affected by the longer wavelengths used for multiphoton excitation. As an example, Brakenhoff et al. found that the axial resolution worsened by 5 μm in comparison to a point-scanning microscope, which had an axial resolution of 1 μm [5]. A very clever solution is temporal focusing, as proposed in [6]. By controlling the temporal profile of the pulse, which is compressed as it propagates through the sample, reaching its shortest duration at the focal plane, and then stretches again as it propagates beyond it, an axial resolution as low as 1.5 μm has been obtained. Clearly, this approach requires proper dispersive optics and careful control of the pulsewidth, depending upon the sample under investigation.

In this work we propose the design and the practical realization of a line-scanning two-photon microscope in which proper arrangement of readily available optical components makes it possible to achieve a resolution comparable to that of point-scanning microscopes. As will be discussed in the following section, an exhaustive review of the literature and a detailed numerical investigation of line-scanning microscopes have led us to the conclusion that the astigmatism introduced by cylindrical optics plays a major role in the degradation of optical-sectioning capability. As a consequence, if this astigmatism is properly managed by altering the choice and positioning of the lenses, the line-scanning microscope can yield optimized performance.

2. Materials and Methods

The realization of a line-scanning fluorescence microscope commonly relies on the insertion of one or more cylindrical lenses into the set-up of a conventional point-scanning microscope. However, as a consequence of the astigmatism caused by these optical elements, two focal planes are generated at a small distance from each other [7]. In the case of confocal microscopy, this is not an issue, as the confocal aperture rejects all the contributions not coming from the conjugated plane [5]. On the other hand, the two focal planes turn out to be detrimental to the performance of a multiphoton microscope. In fact, nonlinear fluorescence is efficiently generated in both planes and collected by the objective. As no aperture screens out the signal arising from the one focal plane not to be imaged, a large amount of undesired fluorescence is able to reach the photodetector along with the proper signal from the other focal plane. Because of the high intensity of the excitation beam, the region of the sample in between the two focal planes is expected to contribute to the amount of additional background fluorescence to a greater extent when the two planes are closer together. Indeed, an increase in the total generated fluorescence with sample thickness has been observed since the first experimental demonstration of a line-scanning two-photon microscope [4], and a marked improvement in the axial resolution upon using a confocal aperture has been demonstrated [5].

In order to assess the accuracy of our arguments, we performed a numerical simulation of the propagation of the excitation beam in a typical line-scanning microscope via ABCD matrix analysis. The parameters that we chose originate from the aforementioned literature. The results are summarized in Figure 1, where both a three-dimensional rendering of the light beam and a plot of the intensity versus the displacement from the plane to be imaged are shown. Along the propagation direction, the light beam is focused along a first desired line in the correct plane and along a second undesired line which is perpendicular to the previous one and lies at a very close distance. As a consequence, the intensity displays two peaks of the same order of magnitude so that the background fluorescence excited by the light beam focused on the second focal plane is of the same order of magnitude as the signal fluorescence excited by the light beam focused on the first plane. Moreover, the light beam exhibits a non-negligible intensity in the whole sample thickness of the two peaks combined, which can give rise to further background fluorescence.

In order to avoid such detrimental conditions regardless of the sample under investigation, the design of the microscope set-up has to be such that the focal distance of one of the two planes is infinite by collimating the beam out of the microscope objective in one of the two transverse directions. This requirement can be fulfilled by means of proper arrangement of the lenses used in the optical system, which has to be properly studied. The optical-sectioning capability of the microscope is thus expected to be preserved, notwithstanding the linear focusing.

As far as the transverse resolution is concerned, a uniform distribution of the excitation intensity along the focal line is a key point. In fact if the intensity varies along the line, the power level necessary to excite a detectable amount of nonlinear fluorescence in the low-intensity regions could possibly saturate the nonlinear absorption in the high-intensity regions, thus worsening the transverse resolution. As the beam emitted by a typical excitation laser source has a Gaussian shape, conventional cylindrical lenses will generate a bell-shaped intensity distribution along the focal line. The most straightforward solution to this issue is the overfilling of the entrance pupil of the microscope objective so that the apodization of the beam will limit the intensity variations. Besides the residual imbalance in the intensity distribution, this approach clearly suffers from inefficient use of laser power. A more efficient and high-performing solution is possible by employing a Powell lens, which is able to convert a Gaussian beam into a uniform intensity line. The design of the microscope will be more complex, but the benefit in terms of performance is appealing.

According to these considerations, we designed and experimentally tested several line-scanning two-photon microscope set-ups. After some promising preliminary results [7], we identified the configuration yielding the best overall results, which is presented and discussed in detail in the following.

The excitation light source is a continuous-wave mode-locked Ti–Sapphire laser (Tsunami, Spectra Physics, Milpitas, CA, USA) emitting a train of pulses at a repetition rate of 82 MHz. The wavelength can be tuned in the range from 750 nm to 850 nm, and the pulsewidth is 110 fs. The output beam has a Gaussian shape with a full-width at half-maximum diameter of 2 mm and is magnified by a factor of 1.5 by a Keplerian telescope before entering the microscope set-up, which is schematically shown in Figure 2a. The magnification factor is chosen so as to meet the specifications of the Powell lens on which the beam impinges right after. This Powell lens has a fan angle of 20° and yields a flat-top profile in one of the two transverse directions labeled as the x direction. The beam emerging from the Powell lens is thus highly divergent in the x direction and is subsequently recollimated by means of a 25 mm focal length cylindrical lens, thus making up an effective Galilean expander.

The beam then impinges on a single-axis scanning galvanometer mirror providing deflection along the perpendicular direction labeled as the y direction at a maximum angle of 25°. A 50 mm focal length cylindrical lens acting in the y direction and a confocally arranged 100 mm focal length spherical lens relay the linear beam, with the desired uniform intensity distribution, on the entrance pupil of the microscope objective after transmission by a dichroic beam splitter. The objective is confocally arranged with respect to the spherical lens, so that the emerging beam is collimated in the x direction. In this way the length of the line-shaped beam is kept fixed along the propagation inside the sample, thus allowing precise control of the intensity, making it possible to set the input power at the proper level required to excite fluorescence just in the one focal plane lying at a finite distance [8]. The behavior of the beam in the x and y directions is shown in Figure 2b,c.

We employ oil-immersion objectives with specifications depending upon the sample to be investigated, most notably the required field of view. The two most used are a 100x objective (UPlanFLN, Olympus, Tokyo, Japan) and a 60x objective (CFI Plan Apochromat Lambda, Nikon, Tokyo, Japan). The sample is mounted on a three-axis translation stage for alignment purposes. In particular, a piezoelectric motor driven by a closed-loop control system with a 20 nm resolution can be used for providing the axial displacement required for three-dimensional imaging, as is often carried out. However, our microscope set-up allows for a different solution for optical sectioning, which turns out to be more efficient. As can be seen in Figure 2b,c, axial displacement of the cylindrical scan lens away from the confocal position will have no effect on the collimation of the excitation beam in the x direction, but will cause an imbalance in the telescope that it makes up with the spherical tube lens in the y direction. As a consequence the axial position of the focal plane will shift according to the amount of displacement of the cylindrical lens. Quantitatively, the focal line will lie at a distance d away from the back of the microscope objective, given by

$$d=f_{MO}-{\left({\displaystyle \frac{f_{MO}}{f_{SL}}}\right)}^2∆$$

(1)

where f_SL is the focal length of the spherical lens, f_MO is the focal length of the microscope objective, and Δ is the axial displacement of the cylindrical lens. As f_MO is definitely smaller than f_SL, the resolution of the axial translation of the focal plane is proven to be better than the resolution provided by the device moving the lens.

The fluorescence radiation emitted following the two-photon absorption process is collected by the same objective used for excitation and then reflected by the dichroic beam splitter. A set of high-pass filters can be used for stronger extinction of undesired radiation, which is mainly due to backscattering of the excitation beam. The beam is finally imaged in real time on a CCD camera (CoolSNAP EZ, Photometrics, Tucson, AZ, USA) using a spherical lens with a focal length of 200 mm. The images are acquired by open-source μManager software [9], which can also be employed for their post-processing. The frame readout of the camera sets the limit to the imaging speed of the microscope set-up to several tens of images per second, with the actual number depending upon the region of interest and the binning factor. However the optical set-up makes it possible to achieve higher acquisition rates up to 350 frames per second, as demonstrated by some preliminary experimental tests that show no degradation of image quality.

The average power of the beam entering the microscope set-up is around 300 mW, with the precise value depending on the sample under investigation. The overall loss of the set-up is measured to be about 24%, with the main contribution arising from apodization of the beam after the Powell lens in order to trim the two intensity peaks at the ends. As a consequence, the pulse energy at the sample is about 2.8 nJ, corresponding to a peak intensity in the range of 7 to 12 GW/cm² according to the used objective.

3. Results

As a first test of the performance of our microscope set-up, we measured the axial resolution by recording the signal emitted by a fluorescent polymer microsphere as a function of the axial coordinate. The diameter of the microsphere was 1 μm, and the scanning galvanometer mirror was held at a fixed angular position. From the results shown in Figure 3, the axial resolution is estimated to be 1.2 μm, which is better than the 1.5 μm value reported in [6] for a line-scanning temporal focusing microscope, the 1.7 μm value reported in [10] for a multiline-scanning temporal focusing microscope, the 1.39 μm value reported in [11] for a multiline-orthogonal-scanning temporal focusing microscope. Further, for the sake of comparison, we also evaluated the axial resolution of a point-scan microscope that we set up as follows: the Powell lens and the first cylindrical lens were removed, a second galvanometric mirror was added in order to possibly provide deflection along the x direction, although this was not required for this measurement, and the second cylindrical lens was replaced by a spherical lens of the same focal length. The measured value was 1.1 μm which proves the effectiveness of the design of our line-scanning microscope.

We then investigated several epidermis samples, both healthy and affected by tumors, in order not only to test the capability of real-time image acquisition of our microscope, but also to prove the feasibility of its practical application. In Figure 4, images of different skin cancer forms (a–c) and of a healthy nevus (d) are shown. These images were collected at a rate of 30 frames/s by exploiting the autofluorescence of skin. For the sake of comparison, the intensity level of each pixel of all four images was normalized to the maximum level recorded. Even if the images were not subjected to any kind of post-processing technique, the quality was good.

Finally we show the optical-sectioning capability of our microscope using the set of frames in Figure 5. The sample is a pollen grain with a 60 μm diameter. The images were acquired at different depths in the grain by varying the axial position of the cylindrical scan lens as previously discussed. Also in this case, the images were not post-processed. The different sections of the pollen grain are clearly visualized, thus providing an ideal basis for three-dimensional reconstruction of the sample under investigation. Moreover, we used this set of images for an evaluation of the signal-to-noise ratio, defined as the ratio of the mean signal value to the standard deviation of the background. The retrieved value is 6.8.

4. Discussion

As a whole, the set of results that we have presented confirm the viability of our approach. We have proven both theoretically and experimentally that a line-scanning microscope is able to overcome the issue of weaker focusing of the excitation beam with respect to a point-scanning system, provided that the proper choice and arrangement of the cylindrical optical components is accomplished. Our results compare favorably to the performance of microscopes employing temporal focusing, which is a well-established technique for increasing imaging speed. It is worth pointing out that a conventional point-scanning microscope can be easily and readily converted into a line-scanning system according to our methodology, with no need for dispersive optical components and high-pulse-energy laser sources, which are required for wide-field temporal focusing systems. Currently further optimization of the set-up is being investigated in order to increase the resolution and the acquisition speed of the microscope.