1. Introduction

High resolution imaging and precise localization of two different structures in biological specimens are of major interest in present and future investigations in biomedical sciences. In particular, correlative or colocalization experiments are widely used to study the function and spatial organization of nearby cellular structures of a specific organism with high resolution. Correlative microscopy techniques combine functional analysis with high resolution imaging [1,2], e.g. by combining fluorescence microscopy with transmission electron microscopy (TEM). This approach enables structural analysis with superior spatial resolution compared to what is feasible with conventional optical methods. However, the drawbacks of this correlation method lie in the difficult and distinct sample preparations for two different imaging techniques as well as in superimposing two images referenced to each other relying on an external marker [1]. A way to overcome this issue is to replace TEM with a different technique that is easier to combine with fluorescence microscopy. Thus, a new approach is the use of super-resolution microscopy in correlative microscopy to reveal structures of interest in super-resolution. Super-resolution microscopy methods like stochastic optical reconstruction microscopy (STORM) [3], photoactivated localization microscopy (PALM) [4] or stimulated emission depletion microscopy (STED) are able to reveal features on the nanometer scale.

To investigate two different structures in fluorescence microscopy, typically two separate excitation lasers are used, and if STED microscopy is also applied, at least three lasers are involved. Without a signal that is generated pairwise by these lasers, the colocalization of the probing volumes for different imaging modalities is not guaranteed. Hence, an approach to ensure the perfect overlap of two lasers while addressing multicolor excitation of different fluorophores is two-color two-photon microscopy (2C2P) [5,6] as an extension of two-photon (2P) excited fluorescence [7]. 2C2P enables effective excitation of two or more different fluorophores over a wide spectral range. Furthermore, extending nonlinear microscopy to super-resolution, techniques like 2P-STED have been developed [8–10] and constitute a powerful tool in imaging beyond the diffraction limit while taking advantage of less scattering and higher tissue penetration depths [11]. Accordingly, those techniques are used in deep tissue imaging reaching depths of hundreds of micrometers, e.g. applying them to image mouse brain tissue [12]. Despite the synergetic advantages of combining the aforementioned microscopy techniques, such an experiment has not been reported yet.

In this work, we developed a novel correlative two-color two-photon excitation STED (2C2P-STED) microscopy technique that enables the precise optical localization of two differently labeled structures in biological specimens with super-resolution. Hence, our imaging approach features the advantages of colocalization microscopy on the nanometer scale. Imaging in both, confocal- and super-resolution, occurs only if the two lasers overlap spatially at the focal volume. This approach leads to an intrinsic alignment of the both images acquired with two different modes. The key innovation of 2C2P-STED is the combination of two lasers operating in the near-infrared (NIR) wavelength region, specifically a ps-diode laser with switchable pulse durations and a fs-fiber laser providing a fixed pulse duration. The diode laser operates in two configurations: a short pulse mode providing pulses around 37 ps and a long pulse mode providing pulses around 538 ps. If the short pulse of the diode laser is combined with the fs-pulse of the fiber laser, 2C2P excitation of a suitable fluorophore takes place (Fig. 1(a)). In order to switch to 2P-STED imaging, the long pulse of the diode laser is phase modulated by a spatial light modulator before combining it with the fs-pulse of the fiber laser (Fig. 1(b)).

 

Fig. 1. Principle of the 2C2P-STED technique. (a) Simplified Jablonski diagram of the two-color two-photon excitation by absorption of two photons with different wavelengths ${\lambda _1}$ and ${\lambda _2}$. The sum of the photon energies excites the molecule, which in turn requires the spatial overlap of both laser foci. 2C2P is used for excitation of a counterstain in a biological specimen. (b) Jablonski diagram showing the two-photon absorption of photons with wavelength ${\lambda _2}$ and the stimulated emission induced by a photon with wavelength ${\lambda _1}$. This leads to an effective point spread function (PSF) beyond the diffraction limit, if a donut-shaped intensity distribution is applied. 2P-STED is applied for imaging the sub-structure of interest. (c) Despite switching between 2C2P (a) and 2P-STED (b), the laser alignment is well preserved. The overlay of the counterstain and the sub-structure image reveals the spatial correlation of the different components.

Download Full Size | PDF

The acquired diffraction limited and super-resolution images are self-referenced to each other, since both lasers are involved in both imaging processes (Fig. 1(c)). Image acquisition is realized frame-by-frame by switching between the two different modes without modifying the beam propagation path, thereby preserving the initial alignment.

To the best of our knowledge, we demonstrate 2C2P-STED for the first time using the Gram-positive bacterial strain Streptococcus pneumoniae TIGR4 expressing pilus type-1. As the substructure of interest, the pilus type-1 RrgB protein is labeled with AlexaFluor594, whereas the cellular shapes of the bacteria are counterstained using ATTO425. The pili protruding from the cell surface are spatially organized in sub-micron proximity and represent a virulence factor promoting the pneumococcal pathogenicity [13,14]. Additionally, we evaluate the maximal resolution in the different imaging modalities by means of numerical calculations. Finally, we characterize the two-color two-photon absorption by means of sum-frequency generation (SFG) and fluorescence microscopy and discuss the impact and possible applications of this technique.

2. Methods

2.1 Experimental setup

A fundamental concept and major advantage of 2C2P-STED is that no changes in the optical beam path are necessary to switch between the two imaging modalities, i.e. 2C2P excited fluorescence and 2P-STED. Therefore, the entire microscope setup is controlled electronically. A schematic drawing of the setup is shown in Fig. 2(a). Two different laser systems are used to combine the nonlinear excitation scheme with stimulated emission depletion. An ultrashort pulsed fiber laser “FEMTO” (FemtoFiber Dichro Design, TOPTICA Photonics AG) is used for two-color two-photon absorption as well as for two-photon absorption. The center wavelength is at 1034 nm with a pulse duration of 95 fs at a repetition rate of 80 MHz. The second laser source is a fiber amplified diode laser “DIODE” with an emission wavelength of 775 nm (Katana 08 HP, OneFive). A major feature of this laser is the operation in two different modes referring to the pulse duration. The pulse duration can be switched electronically between 37 ps “PICO” used for two-color two-photon absorption and 538 ps “STED” used for stimulated emission depletion. In order to synchronize both lasers, a trigger signal provided by the FEMTO laser controls the subsequent emission of the DIODE laser. In addition to the synchronization, an exact temporal alignment of the involved laser pulses at the focal plane is crucial. Therefore, the trigger signal is electronically delayed by a delayer module “PSD” (Picosecond Delayer, Micro Photon Devices), which allows adjustment of the delay of the trigger signal with a resolution of 10 ps in the range of 0 to 50 ns. The virtual excitation wavelength for 2C2P is calculated by ${\lambda _{2C2P}} = ({\lambda _1}{\lambda _2})/({\lambda _1} + {\lambda _2})$ and for 2P by ${\lambda _{2P}} = {\lambda _2}/2$. In our case, ${\lambda _{2C2P}}$ and ${\lambda _{2P}}$ are 443 nm and 517 nm, respectively.

 

Fig. 2. (a) Schematic drawing of the experimental setup: DIODE: 775 nm laser, FEMTO: 1034 nm laser, PSD: Picosecond delayer, M: Mirror, T1: Telescope, PBSC: Polarization beam splitter, SLM: spatial light modulator, HWP1/HWP2: zero-order half-waveplates, D1, D2; D3: dichroic mirrors, QWP: achromatic quarter-waveplate, G: resonant-galvo scanner, SL/TL: scan lens/tube lens, OL: objective lens, C: Condensor lens MF: Multiphoton filter, FB/FR/FF: Bandpass filter, PMT: photomultiplier tubes, MPM: Multiphoton microscope. (b) Visualization of the different imaging modalities. 2C2P Imaging: SLM phase is set to a homogeneous gray value of 127 (8-bit), the DIODE laser is set to PICO mode (37 ps), and the time delay is set to zero. STED imaging: A vortex phase from 0 to 2π is applied on the SLM, the DIODE laser is set to STED mode (538 ps), and a time delay of approximately 250 ps is induced.

Download Full Size | PDF

In order to apply super-resolution microscopy, a reflective spatial light modulator “SLM” (X13138-02, Hamamatsu Photonics) designed for 775 nm is used for wavefront modulation of the STED laser pulse [15,16]. The underlying principle of a SLM is a liquid crystal display (LCD). The SLM converts an uploaded gray scale (8-bit) image to a phase shift in the range of 0 to 2π at the respective SLM pixel. Pixel resolution of the SLM is 1272 × 1024 (pixel pitch 12.5 µm). To achieve the best performance of the phase modulation by the SLM, horizontal polarization of the incident laser beam is ensured using a polarization beam splitter cube “PBSC” (PBS102, Thorlabs). The beam size and divergence of the DIODE laser is adjusted using a variable telescope “T1” (1-3x VBX, NIR, Edmund Optics).

To adjust the linear polarization state of each laser beam, zero-order half-waveplates for 775 nm “HWP1” (460-4220, Eksma Optics) and for 1034 nm “HWP2” (460-4208, Eksma Optics) are used. The laser beams are combined at the dichroic mirror “D1” (Semrock) and guided to a home-built multiphoton microscope “MPM” based on a commercially available microscope stand (Nikon Ti2 eclipse, Nikon). A resonant-galvo scanner “G” (Thorlabs) is used for laser-scanning. An achromatic quarter-waveplate “QWP” (AQWP10M-980, Thorlabs) placed in front of the laser scanner converts both incident laser beams into circular polarization. The combination of half-waveplates and an achromatic quarter-waveplate allows for precise adjustment of a circular polarization at both wavelengths. The two lasers propagate through the telescope formed by the scan lens “SL” and tube lens “TL” to the objective lens “OL” using the sideport of the Nikon stand. An oil immersion objective lens (CFI Apochromat, Nikon) with NA = 1.49, 100x magnification and chromatic correction in the range of 435 nm – 1064 nm is used for focusing.

Fluorescence from the sample is collected by the same objective lens in backward direction before being separated from the illumination beam path by means of the dichroic mirror “D2” (FF825-SDi01, Laser2000). Detection is carried out in a non-descanned configuration using InGaAsP photomultiplier tubes “PMT” (Thorlabs). The signal is separated into the blue and red spectral range by the dichroic mirror “D3” (HC 735 LP, AHF Analysetechnik). The filters “FB” (F76-594, AHF Analysetechnik) and “FR” (AT 600 LP, AHF Analysetechnik) clean up the fluorescence signal in the respective spectral range. Two multiphoton filters “MF” (F39-745, AHF-Analysetechnik) block the excitation and depletion wavelengths to ensure low background signal from the laser light.

A third PMT in forward direction is used for measuring the pulse duration of the DIODE laser in PICO mode by cross-correlating the signal between the FEMTO and the PICO laser using SFG. The width of the cross-correlation is given by ${\tau _{cc}} = \sqrt {\tau _{FEMTO}^2 + \tau _{PICO}^2} $. The FEMTO pulse width of about 100 fs at the sample plane is much shorter than the PICO pulse width. Therefore, the cross-correlation directly gives the temporal pulse shape of the PICO laser at the focal plane ${\tau _{cc}} \approx {\tau _{PICO}}$. The substrate used for SFG is iron-(III)-iodate (Fe(IO₃)₃), which exhibits a strong nonlinear coefficient [17]. Furthermore, this sample is suitable for SFG microscopy due to its chemical stability and the simple preparation: for imaging purposes, the dry, crystalline powder is sandwiched between a high precision cover slip and a microscopic slide and sealed with nail polish. The signal emitted by the substrate is collected by the condenser lens “C”. A multiphoton Filter (F39-745, AHF-Analysetechnik) blocks the fundamental wavelengths, and a bandpass filter “FF” (450/50, AHF-Analysetechnik) separates SFG signal from the second-harmonic signals produced by each laser.

The experiment is controlled by a self-written LabView (LabView, National Instruments) program, including the SLM control, the temporal delay and the PICO/STED laser control. The ThorImage (Thorlabs) software is used for imaging, and data analysis is performed using ImageJ [18] and Matlab (The MathWorks).

In order to switch between the two different imaging modalities, a few parameters must be changed: the pulse width and the phase modulation of the DIODE laser as well as the time delay between the FEMTO and DIODE. The respective settings for the imaging modalities are visualized in Fig. 2(b). In order to perform 2C2P imaging, no phase modulation of the DIODE laser in PICO mode is necessary. Therefore, an image with a homogeneous gray value of 127 (8-bit) is displayed on the SLM. This leads to an unmodulated and almost lossless reflection of the 775 nm laser beam. Additionally, the pulse width is set to PICO mode, i.e. to 37 ps, and the time delay between the FEMTO and the PICO pulse is set to zero using the delay module. Due to the short pulse duration and the exact temporal overlap, a good excitation efficiency is achieved.

For the two-photon STED imaging, the DIODE laser is set to STED mode with a pulse duration of 538 ps. For good depletion efficiency, the STED pulse is delayed by approximately 250 ps in respect to the FEMTO pulse (peak to peak). Since the DIODE laser is used in the long pulse mode, direct two-photon excitation by the STED laser is avoided and nonlinear photodamage is reduced. The STED laser is modulated by a vortex phase from 0 to 2π by the SLM. This leads to a donut-shaped intensity distribution in the focal plane used for depletion. Image acquisition is realized frame-by-frame while alternating between the two different imaging modalities. Moreover, every parameter is set electronically and no changes in the optical beam path are necessary. This keeps the preset alignment and polarization preserved during the switch between the imaging modalities.

2.2 Streptococcus pneumoniae sample preparation

The sample of our proof-of-principle experiment is the Gram-positive bacterial strain Streptococcus pneumoniae TIGR4 expressing pilus type-1. Type-1 pili are hair-like filaments attached at the bacteria surface that are often located at the division plane or encircling the cell [19]. They show a diameter of approximately 10 nm and can be over 1 µm long. Counterstaining is applied on the pneumococcal cellular shapes, whereas the pili are the sub-structures of interest. Streptococcus pneumoniae TIGR4 wildtype strain was grown on blood agar plates for less than 12 hours at 37°C in candle jars. The next day, 5 ml Todd-Hewitt Broth (CarlRoth) supplemented with 5% yeast extract (THYE) was inoculated from blood agar plates with starting optical density (OD) of 0.01 and grown to early log-phase (OD 0.1–0.2) at 37°C without shaking. Cultures were then supplemented with 15% Glycerol and stocks were stored at -80°C. The following staining experiments were started from these stocks. For antibody staining, 1 ml of THYE was inoculated with 50 µl cell material from glycerol stocks and grown for 3 hours at 37°C without shaking. Cells were then centrifuged at 3200 g for 5 min at 4°C and washed 3 times in 1 ml phosphate-buffered saline (PBS). After resuspension, cells were fixed and subsequently washed in PBS. Cells were then resuspended in 1 ml PBS and incubated 1:500 with anti pilus type-1 backbone antibody (α-RrgB) rabbit serum for 45 min at 37°C without shaking, followed by 3 washing steps in 1 ml PBS. Subsequent staining with AlexaFluor594-conjugated α Rabbit secondary antibody (1:250) (ThermoFischer) was carried out at 37°C for 45 min without shaking. After 3 subsequent washing steps with PBS, cells were counterstained using 10 µM free ATTO425-maleimide (AttoTEC GmbH) dye in PBS for 30 min, followed by 3 x washing steps in 1 ml PBS. Cells were then spotted on 170 µm high-precision coverslips (CarlRoth) and allowed to air dry for 15 min, followed by mounting with ProLong Diamond (ThermoFisher) on glass slides suitable for fluorescence microscopy.

2.3 Numerical calculations of the effective point spread functions

In biological imaging, resolution is a very important parameter. Theory of diffraction predicts a strong dependency of resolution on the wavelength [20]. In nonlinear microscopy, wavelengths in the near-infrared spectral range are widely used, and advantages such as less scattering, less wavefront aberrations and a higher intrinsic axial resolution [11] counteract the disadvantage of a lower resolution. Despite the longer wavelengths, the quadratic dependency [11,21] on the excitation intensity leads to a resolution suitable for most applications. To give an impression of the resolution of each imaging modality used in two-color two-photon STED microscopy, the PSFs of the wavelengths used in this study are determined by numerical calculations.

We calculate the three spatial electric field components $({{E_x},{E_y},\; {E_z}} )$ of a strongly focused laser beam using the vectorial Debye theory [22–26] Eq. (1). In this calculation, we assume circular polarization of the incident laser beams, as it is also applied in the experiment. In Eq. (1), f denotes the focal length of the objective lens and λ the illumination wavelength. The polar angle is given by θ and the azimuthal angle by ϕ. The illumination ratio of the back aperture is given by the factor γ. In the present calculations, double-illumination of the back aperture is considered (γ = 0.5). The maximum polar angle α is given by the numerical aperture $NA = n\sin (\alpha )$ of the objective lens assuming a refractive index of n = 1.5175. The wavenumber is denoted as k and $({x,y,z} )$ are the Cartesian coordinates. The phase factor ${e^{i{\Delta }\alpha }}\; $ describes the spatial phase modulation of the incident laser beam. In the case of STED microscopy, ${\Delta }\alpha = \phi $ applies to create the donut-shaped intensity distribution in the focal plane [27,28]. For two-color two-photon imaging, the phase factor is given by ${\Delta }\alpha = 0$. The integrals are calculated in the ranges of $\theta = 0\; ..\alpha $ and $\phi = 0..2\pi $.

3. Results

3.1 2C2P-STED microscopy of Streptococcus pneumoniae TIGR4 pilus type-1

Following the principle of 2C2P-STED (Fig. 1), we investigate pilus type-1 of Streptococcus pneumoniae TIGR4. First, we acquire a two-color two-photon image using the 2C2P imaging conditions as described in section 2.1 and Fig. 2(b). The cellular shapes of the bacteria are stained with ATTO425. The emitted fluorescence photons acquired with PMT1 (Fig. 2(a)) are related to the scanner position and form the intensity image (Fig. 3(a.2)). In this first approach, the diffraction limited two-photon image of RrgB pilus-1 components (Fig. 3(a.1)) labeled with AlexaFluor594 is simultaneously acquired using PMT2 (Fig. 2(a)). Notably, in conventional laser scanning microscopy, these two images are not referenced to each other, and the quality of localization depends on the quality of the alignment of both lasers. In the 2C2P approach, the signals are self-referenced to each other, since the fluorophore used for counterstaining is excited only in the case that both lasers overlap spatially and temporally. This signal serves as a reference between the two images. In order to obtain a high-contrast image with high localization accuracy, the fluorophore used for 2C2P may only be excited with the combination of both lasers. The requirements of the fluorophores for 2C2P-STED are discussed in detail in section 3.4.

 

Fig. 3. 2C2P-STED of S. pneumoniae TIGR4 and pili type-1. a) Intensity images of the two separate detector channels. a.1) AlexaFluor594 labeled RrgB components with 2P excitation. a.2) ATTO425 labeled cellular shape applying 2C2P excitation. a.3) STED image of AlexaFluor594 labeled RrgB components. b) Merged images acquired with 2C2P and 2P mode in diffraction limited resolution (left) and with STED microscopy (right). Images are shown in false color (Cyan: ATTO425, Red: AlexaFluor594). Here, the pili or subunits of the pili are predominantly localized at the polar and midcell region. c) The magnified 2C2P-STED image of the bacterium (white arrow in (b)) reveals a separated sub-structure of labeled RrgB pilus-1 components along the equatorial plane. d) Intensity profiles for STED and 2C2P fluorescence along the lines that span between the red and green arrows, respectively. RrgB labeled components in the distance of 70 nm can be clearly distinguished. Scale bar a) and b) 1 µm; c) 500 nm.

Download Full Size | PDF

After imaging with diffraction limited resolution, automatically changing the laser and imaging parameters prepares the setup for two-photon STED imaging. In order to apply 2P-STED, the LabView program changes the phase mask at the SLM to a vortex, switches the DIODE laser to the long pulse mode (STED mode) and changes the time delay between the excitation pulse and the STED pulse to approximately 250 ps (see Fig. 2(b) – STED imaging). The long pulse is necessary to avoid direct two-photon excitation by the STED laser itself. In this configuration, the RrgB labeled pilus-1 components are imaged again, but now with sub-diffraction limited resolution (Fig. 3(a.3)). The STED imaging results from detection of the fluorescence signal by PMT2 and forms an intensity image (Fig. 3(a.3)). Here, the background induced by the STED laser itself was subtracted pixel-by-pixel from the raw STED image. For this purpose, an image was acquired under the same conditions as for the STED image but with the difference, that the excitation laser was blocked. The weak background in the raw STED image shows a homogeneous distribution resulting in a signal-to-noise ratio that exceeds 4.5. The overlay of both the diffraction limited and sub-diffraction limited images of pilus-1 with the counterstained cellular shapes reveals the surface localization of pilus-1 RrgB structures (Fig. 3(b)). STED images of pilus-1 at resolution below the diffraction limit have not been described for respective antibody-mediated fluorescent imaging so far. The main advantage of the 2C2P-STED technique is that the image regarding the counterstain (cellular shapes) and the image of the sub-structure of interest (pilus-1) are referenced to each other. In this approach, only the laser and phase modulation parameters are changed, thus the alignment is preserved while switching between 2C2P imaging and 2P-STED imaging. Here, the alignment between the two lasers in 2C2P excitation is ensured with a precision better than the diffraction limit [5], otherwise no signal of the 2C2P excited fluorophore would be detected. Images were acquired taking the average over 60 frames at a scanning rate of 7.4 frames per second, which leads to an acquisition time of approximately 16 s. Image acquisition is realized frame-by-frame. The pulse energy coupled into the back aperture of the objective lens was 375 pJ for the FEMTO in both the 2C2P and 2P excitation. In case of 2C2P excitation, the pulse energy of the PICO laser was 500 pJ. In STED, the pulse energy of the donut-shaped STED laser was 400 pJ.

Figure 3(b) shows clearly that pili can accumulate at the polar regions and the middle part of the cells, potentially colocalizing with the septal region of the bacteria. In particular, RrgB labeled components span over the equatorial plane of the cells and show a thickness of approximately 170 nm to 180 nm. The magnified merged 2C2P-STED image (Fig. 3(c), right) reveals that features in the sub-100 nm range can be clearly distinguished. This is quantified by measuring the distance between separated RrgB labeled components at midcell using a profile plot (Fig. 3(d), left), which shows components well separated with distances in the order of 70 nm. In Fig. 3(d) right, the profile plot reveals that peaks resulting from RrgB components narrow down in the STED image to approximately 88 nm compared to 265 nm in 2C2P imaging, which corresponds to a three-fold improvement in resolution.

3.2 Numerical calculations and achievable resolution

2C2P-STED correlates a diffraction limited image with a sub-diffraction limited image combining 2C2P technique with 2P-STED. Commonly, NIR wavelengths are used for nonlinear microscopy that lead to lower resolution compared to one-photon confocal and STED microscopy operating in the visible spectral range. Therefore, we investigated the theoretically achievable resolution by calculating the intensity distribution – i.e. the effective PSF – in the lateral focal plane (XY-plane) for both imaging modalities considering the wavelengths 775 nm and 1034 nm. These wavelengths are also good representatives for other nonlinear imaging techniques. A commonly used approximation for the resolution in conventional microscopy is to simply divide the excitation wavelength by a factor of two. In practice, the resolution also depends on the sample, the fluorophore and the detection itself, which leads to a slightly decreased resolution in contrast to the bare PSF. In 2C2P imaging, the two lasers are focused and form the effective excitation PSF shown in the right column in Fig. 4. For the 1034 nm laser, the full width at half maximum (FWHM) of the respective one photon excitation PSF is approximately 349 nm. This is almost a third of the wavelength and can be explained mainly by the high numerical aperture of NA = 1.49. In case of the 775 nm, the FWHM of the PSF is 262 nm, which is also almost a third of the wavelength. Values calculated can be reached with the setup described in section 2. The effective excitation PSF is calculated by multiplying each PSF of the involved lasers according to Eq. (3) and results in the FWHM of 215 nm. This value is below the particular PSFs at the wavelengths of 775 nm and 1034 nm, respectively. To put this into perspective, a single photon process in the visible spectral range at 517 nm has an effective PSF width (FWHM) of approximately 170 nm (not shown here). Consequently, compared to one-photon confocal microscopy, the effective PSF in 2C2P excitation is only slightly larger even though the wavelength is doubled. In particular, the effective 2C2P-FWHM is approximately half the wavelength of the virtual excitation wavelength ${\lambda _{2C2P}}$, which is in good agreement with the general approximation described above.

 

Fig. 4. Simulation of the intensity distribution in the XY-plane for 2C2P imaging and STED imaging modality. Upper part (2C2P imaging): Normalized intensity distribution of each singular process with $\lambda = 775\; \textrm{nm}$ and $\lambda = 1034\; \textrm{nm}$ for 2C2P imaging. Here, the effective PSF results as the product of the intensity distributions of each focused laser ${I_{2C2P}} = {I_{1034}} \cdot {I_{775}}$ (right column). Lower part (STED imaging): Two-photon excitation intensity distribution with $\lambda = 1034\; \textrm{nm}$ and donut-shaped intensity distribution used for stimulated emission depletion with $\lambda = 775\; \textrm{nm}$. The full width at half maximum (FWHM) of the effective PSFs in each imaging modality is given in the images. In STED imaging the resolution increases with higher STED laser intensities shown in the lower row in the right column.

Download Full Size | PDF

The intensity distribution in the lateral plane is uniform and circular-shaped due to circular polarization of both lasers. It is worth to point out that the equally oriented circular polarization of both lasers is a critical parameter to reach a uniform and high excitation efficiency of the fluorescent molecules. Furthermore, side slopes that are present in one-photon processes are suppressed in 2C2P excitation.

On the other hand, in STED microscopy the effective PSF results from switching off the fluorescence in the circular outer region of the excitation area [29]. In this work, two-photon STED microscopy using 1034 nm is applied. In 2P excitation, the intensity distribution in the focal volume is of quadratic dependency (Eq. 4) on the focused intensity. The 2P intensity distribution for 1034 nm is shown in the lower part of Fig. 4 (STED imaging). The FWHM of this PSF calculates to about 254 nm and is in the order of half the wavelength given by the virtual excitation wavelength ${\lambda _{2P}}\; $ of 517 nm. Here, the general approximation of the achievable resolution also applies. By means of STED microscopy, the effective PSF narrows down by stimulated emission with $\lambda = 775 \,\textrm{nm}$ in regions where the donut-shaped intensity distribution overlaps with the 2P excitation area.

Considering only the geometric aspect, the resulting FWHM of the effective STED PSF is lower than 150 nm. Notably, this value strongly depends on the stimulated emission cross-section of the fluorophore, which in turn depends on the wavelength and the actual peak power of the STED laser. A general statement about the resolution is therefore not possible; rather, in STED microscopy the resolution is theoretically not limited.

By increasing the STED laser power, the effective PSF can be further narrowed down to values far below 50 nm. A resolution of up to 5.8 nm has been shown in practice [30]. The resolution depends on the STED laser intensity as described in [31]:

3.3 Sum-frequency generation (SFG)

Applying nonlinear excitation, a high photon density in the focal plane is crucial for efficient excitation. Therefore, fs-lasers are usually used due to their high peak power. In our novel approach, we combine a fs- laser with a picosecond laser for inducing nonlinear excitation. In order to both show that the excitation is nevertheless efficient as well as characterize the temporal pulse shape of the PICO laser, a cross-correlation between the FEMTO and the PICO laser is acquired. For this purpose, the SFG signal is measured at different time delays using the sample iron-(II)-iodate. The microscope’s group delay dispersion at 1034 nm is determined to be 3035 fs² [33]. This leads to a pulse duration of the FEMTO laser at the focal plane of approximately 130 fs. Microscopic images of the crystalline structure and the measured cross-correlation signal of both lasers are shown in Fig. 5(a) and (b), respectively. Images are acquired in the range of 500 ps in steps of 10 ps. The cross-correlation signal is evaluated as mean intensity value of the region of interest (ROI) (white box Fig. 5(a), image 3) that is shown in Fig. 5(b).

 

Fig. 5. Cross-correlation of the FEMTO and the PICO laser in steps of 10 ps using SFG. The sample used is Fe(IO₃)₃. a) Microscopic images of the SFG signal at different time delays (images 1-5). The time dependent intensity signal is evaluated in the region of interest (ROI) shown in image 3. Scale bar: 2 µm b) The cross-correlation directly gives the pulse shape of the PICO laser. Inset: Schematic energy diagram of the SFG process. The main peak has a FWHM of 37 ps. The peak (position 3) is defined as “time-zero”.

Download Full Size | PDF

Apparently, the temporal pulse shape of the PICO laser is composed of two parts: a narrow main peak (3) and a broader shoulder (2). The main peak exhibits a FWHM of 37 ps whereas the shoulder is of approximately 100 ps in width. The intensity at time delay of 0 ps (3) is almost three times more than at 50 ps (2). To ensure the best excitation efficiency, the good overlap between the FEMTO and PICO laser is given at the main peak (3), which is defined as “time-zero”. For a strong nonlinear excitation, a high peak power is of major interest. In this experiment, the peak power for the FEMTO laser was 900 W and for the PICO laser 11 W, taking only the main peak into account.

Interestingly, the signal can be scaled using high peak power of the FEMTO laser although with low peak power of the PICO laser. The emitted signal at SFG is at a defined wavelength due to the parametric process and can be clearly separated from other contributions, such as second-harmonic generation (SHG) of both the FEMTO or PICO laser. In this case, there is no signal from the FEMTO laser alone. Therefore, despite low peak powers of the long-pulsed laser, it is possible to reach a good contrast and high SNR. The present experiments show a very good contrast and a high SNR (maximum to minimum intensity) greater than 20. As the situation could be different in fluorescence microscopy, we investigate the two-color two-photon excited fluorescence in the next section.

3.4 Background and signal-to-noise of two-color two-photon (2C2P) excitation

We characterized the nonlinear excitation efficiency of the PICO laser in combination with the FEMTO laser using SFG where the signal can be clearly separated from other contributions. In fluorescence, however, the absorption is spectrally broad and direct two-photon excitation of only one of the lasers could additionally contribute to the fluorescence signal. This unwanted background deteriorates the reference mechanism because the spatial overlap of both laser foci is not mandatory to generate fluorescence. Therefore, background-free 2C2P is critical for colocalization in 2C2P-STED. A suitable fluorophore for 2C2P requires high absorption cross-section at the virtual excitation wavelength ${\lambda _{2C2P}}$, whereas the absorption cross-section for the respective wavelengths ${\lambda _1}$ and ${\lambda _2}$ must be very low. In terms of 2P-STED microscopy, the second fluorophore needs to have a two-photon cross-section for 1034 nm as high as possible, a two-photon cross-section for 775 nm as low as possible, and a sufficient stimulated emission cross-section at 775 nm. These requirements are not different from common 2P-STED.

In order to show the contributions of the FEMTO and PICO lasers and the combination thereof, both the 2C2P and 2P imaging modalities are applied on Streptococcus pneumoniae TIGR4 and pilus type-1 (see section 2B) using the same labelling methods described before. In the first instance, we investigate the excitation by means of 2C2P and 2P absorption considering the respective fluorophore. Figure 6(a) shows a phase contrast image of a bacteria chain consisting of three cells. The same region is imaged in 2C2P excited fluorescence using the PICO and the FEMTO laser (Fig. 6(b)). Images shown here feature a high SNR greater than 15. To achieve a good nonlinear excitation efficiency in 2C2P, the time-delay between the FEMTO and PICO laser is set to time- zero defined in section 3.3. The two-photon image of the S. pneumoniae pilus type-1 is shown in Fig. 6(c). The image features a SNR greater than 10. Notably, the FEMTO laser has the same average power for both the 2C2P and 2P imaging.

 

Fig. 6. a) Phase contrast images of Streptococcus pneumoniae TIGR4. b) 2C2P image of S. pneumoniae cellular shapes. c) 2P image of pilus-1 related RrgB components. d) Energy schemes of potential excitation processes considering 2C2P excitation of ATTO425 at the wavelengths λ₁ and λ₂. Unwanted background via direct 2P excitation can be avoided if no transition using either λ₁ or λ₂ is possible. e) Intensity profile along the orange arrow (shown in b)) with 2C2P and direct 2P excitation of ATTO425. Continuous purple: 2C2P. Dashed green: 775 nm. Dotted red: 1030 nm. Scale bar a) – c) 1 µm.

Download Full Size | PDF

Potential two-photon processes responsible for a decreased contrast are shown in the energy level scheme Fig. 6(d), whereas in the best case for 2C2P excitation, only the sum of both laser photons excites the molecules. Therefore, the 2P contributions of the respective lasers are investigated. In Fig. 6(e) the intensity profile along the cells (orange arrow Fig. 6(b)) in case of 2C2P excitation and direct 2P excitation is shown. Clearly, 2C2P has the strongest signal and the single cells can be discriminated very well. Direct 2P excitation is below 953 counts in each case. In comparison, 2C2P imaging has a maximum intensity of 8616 counts, which leads to a SNR of 9 in case of the PICO laser and 12 in case of the FEMTO laser.

4. Discussion and conclusion

We implemented a novel method combining two-color two-photon microscopy with STED microscopy, called 2C2P-STED. The technique is capable of correlating diffraction limited and sub-diffraction limited images and allows for robust, self-referenced colocalization experiments with super-resolution.

2C2P-STED microscopy has been achieved by combining two lasers (775 nm and 1034 nm), each serving two different functions in the two imaging modalities: the diode laser contributes to the virtual excitation wavelength regarding 2C2P excitation and provides the depletion beam for STED. The fiber laser contributes to the 2C2P excitation as well and provides two-photon excitation in STED. Only the combination of the different lasers excites the first fluorophore by 2C2P. This leads to a self-referenced two-color fluorescence image with diffraction limited and sub-diffraction limited resolution. Here, imaging is realized frame-by-frame by switching between the two different modes. The switching could take place at shorter intervals, i.e. pixel-by-pixel or line-by-line, in particular if a galvo-galvo scanner or a piezo scanner is applied.

In our experiments, we demonstrated the feasibility of 2C2P-STED on Streptococcus pneumoniae TIGR4 expressing pilus-1. We show the strong two-color two-photon excitation with low background of the bacteria labeled with ATTO425. Surface filaments of the bacteria (pilus-1) were immunostained with AlexaFluor594 and imaged with 2P-STED microscopy. With super-resolution, pilus-1 is investigated beyond the optical diffraction limit and reveals the pilus-1 surface localization in great detail at the polar regions and the middle part of the bacteria. Here, we achieved a resolution of approximately 70 nm. Pili are involved in host interaction and invasion [13,14]. In future investigations, it is crucial to understand the function of the pili as well as factors influencing the pili surface localization for a better understanding of their role as virulence factor. High localization precision and resolution are mandatory to study pneumococcal pilus type-1 organization in detail and will provide novel data of Gram-positive pilus biology. 2C2P-STED combines localization precision and resolution, while intrinsic alignment of the involved lasers is ensured in each case. With this method, influences of e.g. cellular growth state or the capsule formation status on the pili organization can be investigated further.

According to the numerical calculations using the vectorial Debye theory, we found that with the laser wavelengths of 1034 nm and 775 nm used in this work, a resolution of 215 nm in 2C2P excited fluorescence can be achieved. In 2P-STED microscopy the resolution is in principle unlimited [30,31]. The numerical formalism described here can be used to determine the proper selection of a suitable wavelength combination to achieve the best resolution necessary for the respective application.

We characterized the temporal shape of the picosecond pulsed excitation laser by applying SFG using iron-(II)-iodate. Thereby, we found the optimal time delay for 2C2P imaging. In our experiments, we demonstrated SFG using the particular laser combination with pulse durations of 130 fs and 37 ps at the focal plane and reached a SNR greater than 20, which allows for high contrast imaging.

In conclusion, a major advantage of 2C2P-STED is that images from both modalities are intrinsically aligned, since both imaging processes only occur where both lasers overlap spatially. Furthermore, only two lasers are necessary for imaging a two-color stained specimen in 2C2P excited fluorescence and STED. The spatial light modulator used in 2C2P-STED would allow for compensation of wavefront errors [34] to implement adaptive optics or extend the method to 3D-STED by changing from a vortex phase mask to a π-step phase mask [35,36]. The use of near-infrared wavelengths in this method results in reduction of photodamage and less scattering in tissue [8]. Due to the comprehensive electronically approach of 2C2P-STED, the technique can be combined with Gated-STED [37] to further enhance the image quality and reduce photobleaching. The nonlinear nature of the method allows for further combination of other techniques, such as polarization-resolved two-photon imaging enabling sub-diffraction contrast [38].

In future work, fluorescent dyes conjugated to antibodies can be replaced by nanobodies, fluorescent proteins or any different labeling strategies suitable for fluorescence and STED microscopy, to enable e.g. live cell imaging. If the technique is combined with a broadband spectral laser source, e.g. a supercontinuum source or a tunable Ti:sapphire laser, a huge variety of fluorescent markers can be used. This will optimize the method by careful selection of fluorophores and laser wavelengths regarding excitation and depletion efficiency. Additionally, as demonstrated in this work, the method is not limited to fluorescence applications in general. For example, 2C2P fluorescence can be replaced with SFG for intrinsic colocalization.

In summary, 2C2P-STED has significant potential as a robust nonlinear imaging modality in multicolor super-resolution microscopy with a particular applicability in colocalization experiments using NIR wavelengths, such as those used in deep tissue imaging.