Recently, many resolution enhancing techniques are demonstrated, but most of them are severely limited for deep tissue applications. For example, wide-field based localization techniques lack the ability of optical sectioning, and structured light based techniques are susceptible to beam distortion due to scattering/aberration. Saturated excitation (SAX) microscopy, which relies on temporal modulation that is less affected when penetrating into tissues, should be the best candidate for deep-tissue resolution enhancement. Nevertheless, although fluorescence saturation has been successfully adopted in SAX, it is limited by photobleaching, and its practical resolution enhancement is less than two-fold. Recently, we demonstrated plasmonic SAX which provides bleaching-free imaging with three-fold resolution enhancement. Here we show that the three-fold resolution enhancement is sustained throughout the whole working distance of an objective, i.e., 200 μm, which is the deepest super-resolution record to our knowledge, and is expected to extend into deeper tissues. In addition, SAX offers the advantage of background-free imaging by rejecting unwanted scattering background from biological tissues. This study provides an inspirational direction toward deep-tissue super-resolution imaging and has the potential in tumor monitoring and beyond.
Resolution and penetration depths are two key factors in far-field optical imaging of biological tissues. Far-field resolution improvement has been addressed by several innovative approaches, including stochastic localization,1–3 deterministic scanning,4,5 and structured illumination techniques.6,7 The examples of a stochastic approach are photo-activated localization microscopy (PALM)2 and stochastic optical reconstruction microscopy (STORM),3 which are based on wide-field imaging schemes, and hence lack the ability of optical sectioning, in turn ineffective for deep tissue applications. For deterministic resolution enhancement approaches, such as stimulated emission depletion (STED)5 microscopy and ground state depletion (GSD)8 microscopy, they are based on confocal detection and thus provide optical sectioning for tissue imaging. However, these approaches require a spatially modulated depletion beam which distorts easily when imaging deep in tissues. The suggested imaging depth of a commercially available STED microscope is less than 10 μm. With the aid of aberration-correction optics, STED microscopy can achieve three-fold resolution enhancement, which is around 60 nm at a depth of 120 μm,9 but its resolution rapidly drops at deeper regions. Structured illumination microscopy (SIM), though based on wide-field imaging, exhibits the capability of optical section. Nevertheless, similar to STED, it requires spatially modulated illumination that gets distorted easily in deep tissues.7 Although some of these techniques used two-photon excitation to improve the penetration depth, the resolution degrades significantly in deep tissues due to the wavefront distortion.10,11 Attempts of combining selective plane illumination microscopy (SPIM) or light sheet microscopy with various wide-field super-resolution techniques, such as SIM,12 STORM,13 and stochastic optical fluctuation imaging,14 have been made for high resolution three dimensional (3D) imaging of tissue structures. However, the limitation remains for imaging in highly scattering samples due to the wide-field detection scheme in all SPIM techniques; hence, they are applied mostly to transparent samples such as zebrafish. Thus, a major challenge in tissue imaging is the degradation of resolution due to increased light scattering/aberration with imaging depth.15,16
We have developed another novel deterministic approach of resolution enhancement: saturation excitation (SAX) microscopy which is based on saturation of emission and temporally modulated excitation beam to extract the nonlinear components.17 Compared to super-resolution techniques that are based on spatial modulation, temporal modulation exhibits less degradation with imaging depth, so this technique should be promising for deep tissues high resolution imaging. SAX microscopy has been applied to the observation of actin filaments of HeLa cells in a 3D matrix stained with the ATTO488 phallodin fluorescence dye as deep as 40 μm.18 Two-photon fluorescence SAX imaging has been reported to observe a 3D tissue sample at a depth of 100 μm.19,20 However, the resolution enhancement by fluorescence SAX is limited to less than two-fold due to low nonlinearity involved with fluorescence saturation. The long excitation wavelength of two-photon SAX degrades the resolution even further. In addition to that, deep-tissue SAX imaging requires more excitation power to achieve fluorescence saturation and can cause photobleaching of the fluorophore.18,21
Compared to fluorophores, scattering from plasmonic nanoparticles are bleaching unlimited and possess much larger cross sections. We have recently demonstrated that scatterings from various plasmonic nanostructures are saturable22,23 and provide much better resolution enhancement when combined with SAX or other deterministic approaches.22,24 Among different plasmonic materials, gold nanoparticles are more favorable due to their non-toxicity toward biological samples. For instance, these particles had been successfully applied for long-term tumor imaging.23,25–28
In this work, we demonstrate three-fold resolution enhancement with plasmonic SAX microscopy at a depth of 200 μm inside a scattering biological tissue, and the enhancement factor maintains throughout different depths. The imaging depth is only limited by the objective working distance, and the plasmonic SAX detection is promising to extend into deeper tissues. It provides an attractive background-free feature since nonlinear scattering mainly arises from plasmonic particles, not from tissues. The merit of this study lies in that it reveals, for the first time, the potential of plasmonic SAX microscopy in enhancing resolution inside tissues at a depth that typical super-resolution techniques cannot achieve. We anticipate that our study will trigger more investigations on not only applications of plasmonic SAX microscopy but also innovative methods to achieve deep-tissue resolution enhancement.
The basis of SAX microscopy is saturation of emission (fluorescence or scattering) due to nonlinear behavior of the target specimen. For a focused laser illumination, saturation starts from the center, where the intensity is strongest, while the periphery remains unsaturated. The key to improve resolution by SAX microscopy is to spatially extract the saturated part of the emitted radiation, and the trick is to add temporal modulation into the excitation beam. In the unsaturated region, i.e., the periphery of a focus, the emitted signal linearly follows the modulation frequency. In the saturated region, i.e., the center of a focus, the nonlinearity gives rise to signals in different harmonics of the modulation frequency. By selectively extracting the harmonic signals with the aid of a lock-in amplifier (LA), the effective point spread function (PSF) is reduced in all three dimensions (3D), leading to 3D resolution enhancement.17,22 Typically, the resolution enhancement is better with higher order harmonics.
The setup used in this experiment is a standard upright microscope with home-built modulation and scanning and detection units, similar to what we have reported previously.22,29 A simplified scheme of the setup is presented in Fig. 1(a). Briefly, the system uses a continuous wave laser source of wavelength λ = 561 nm (Cobolt Jive™ 561 nm, Cobolt, Sweden). This wavelength matches the plasmonic resonance peak of 80 nm gold nanospheres (GNSs).4 A beam splitter splits the incident light into two arms before transmitted through two acoustic optical modulators (AOMs) (AOM-40, IntraAction, IL) individually. The two first-order diffraction beams out of the AOMs were interfered to generate a sinusoidally modulated illumination beam at 10 kHz frequency. The modulated beam was then raster scanned by a pair of galvanometer-mirrors and focused on the sample by an objective lens (UPlanFL 100×/1.30 Oil, Olympus, Japan), whose working distance is 200 μm. Another beam splitter in the reflection path directs the GNS scattered light to the photomultiplier tube (PMT, H5783-04, Hamamatsu, Japan) through a confocal detection system.
If the scattering signal was saturated, it carried not only the fundamental modulation frequency f = 10 kHz but also high-order harmonics 2f, 3f, etc. The electric output of the PMT was then fed to a lock-in amplifier (SR830, Stanford Research Systems, USA) to filter out the harmonic components. A photodetector that records the fundamental modulation frequency provides a reference to the lock-in amplifier. A computer system then processes individual harmonic signals to form two-dimensional scanning images in synchronization with the scanner.
Chicken breast test tissue samples were collected from freshly slaughtered chicken that was purchased commercially. The tissues were sliced in thicknesses of 50 μm, 100 μm, 150 μm, and 200 μm with a Leica CM1950 clinical cryostat microtome slicer and were collected on cover glasses. Plasmonic GNSs, of diameter 80 nm, were attached onto the tissue surface by keeping them unperturbed for 5 min. The unattached excess particles were washed away by deionized water. After the attaching process, the tissue sections were immersed in oil to avoid reflection background from the tissue-glass interface and protected by a slide glass in the bottom, as shown in Fig. 1(b). Figure 1(c) demonstrates a typical SAX image of GNSs distributed over muscle fibers formed with 1f harmonic signals, where the sarcomere lines, i.e., the signature of muscle fires, are indicated by a white arrow.
The images were analyzed with Java based image processing software ImageJ, to determine the spatial resolution and contrast [signal-to-background ratio (SBR)]. After importing the raw data, a line profile passing through the center of each particle was drawn with the help of a plot profile plugin in ImageJ. From the profile, the FWHM of the curve was measured to determine its spatial resolution. The contrast was determined by dividing the averaged intensity of central 3 × 3 pixels of each particle versus the averaged intensity of 25 × 25 pixels in the region with only tissues and no particles.
Figure 2 shows the representative SAX images of GNSs, with and without tissues. The first column shows images of particle without tissues, and the second to fifth columns present SAX images of particles embedded in tissues of 50 μm, 100 μm, 150 μm, and 200 μm thicknesses, respectively. The first row demonstrates the images of GNSs demodulated at f = 10 kHz (1f image). Since the 1f image corresponds to the linear components of scattering, the images are overlapped with the scattering background from muscle tissues. Due to the high contrast of GNSs over muscle fibers and the small imaging area, the muscle sarcomeres are not clearly visible in the 1f images here. The second and third rows of Fig. 2 show resolution enhanced images of GNSs formed by 2nd harmonic (2f images) and 3rd harmonic (3f images) signals, respectively. Since tissue scattering disappears in 2f and 3f images, it suggests that tissue scattering exhibits a much smaller nonlinear response. As a result, plasmonic SAX imaging provides an attractive background-free advantage.
It is obvious that the FWHM of every particle is significantly reduced in the 2f and 3f images of Fig. 2. The enhanced resolution is nicely demonstrated by two adjacent particles in the third column, as highlighted by white arrows. These two particles are not resolvable in 1f image but are fully resolvable in higher harmonic images.
To quantify the resolution enhancement by plasmonic SAX microscopy and determine its depth dependence, we have measured the FWHM of GNS PSF for 1f, 2f, and 3f harmonic signals through all different tissue thicknesses. The resolution enhancement factors (REFs) are then defined by the ratio of FWHM of particle images obtained with demodulation frequencies of 1f–2f (1f/2f) and 1f–3f (1f/3f), respectively. Figure 3(a) shows the value of REF with different tissue thicknesses. For 2f signals, the range of resolution enhancement is 1.5 ± 0.16–1.8 ± 0.23, while for 3f signals, the range is 2.3 ± 0.3–3.0 ± 0.6. These values not only correspond well to our previous measurement22 but, most important of all, do not degrade as imaging depth increases, as we expected.
On the other hand, although the REF does not change with thickness, the absolute PSF size does, reflecting the strong scattering and aberration in our sample. The effect is manifested in Fig. 3(b) which shows the PSF size enlargement with tissue thicknesses. The column plot demonstrates the FWHM change of GNS PSFs for the three different harmonics. The FWHM in the 1f images enlarges from 0.56 ± 0.1 μm without any tissue to 1.38 ± 0.2 μm through 200 μm tissues. Similarly, 2f enlarges from 0.34 ± 0.05 μm to 0.81 ± 0.15 μm and 3f enlarges from 0.24 ± 0.04 μm to 0.47 ± 0.15 μm. This PSF enlargement is the result of scattering and multiple refractive-index-mismatch induced aberrations.30,31
From the results in Fig. 3, it is apparent that SAX works for both diffraction-limited and non-diffraction-limited imaging systems. In our previous work, we have demonstrated three-fold resolution enhancement in a diffraction-limited system.22 In this work, the intrinsic aberration in the optical system makes the FWHM of the 1f image larger than the diffraction limit, and the scattering/aberration of biological tissues further enlarges it. Nevertheless, as we demonstrated, three-fold resolution enhancement is achieved in the presence of scattering and aberration, with or without tissues. The underlying reason is because SAX enhances spatial resolution by extracting nonlinear responses, which is more pronounced in the high-intensity area. As long as the PSF remains to be a peak, where intensity is higher at the center, the intensity difference between the center and the edge of the PSF can be magnified by SAX, thus reducing the effective width and leading to enhanced resolution.
To investigate the imaging depth limitation, the signal to background ratio (SBR) is characterized in Fig. 4. As mentioned earlier, one advantage of plasmonic SAX is to provide strong reduction of the background signals. This is demonstrated by the significantly enhanced SBR of 2f and 3f signals versus that of 1f, when tissues exist with GNSs. Typically, the signal strength should decrease with increasing harmonics. Nevertheless, since the nonlinear effect from tissues is much smaller from that of nanoparticles, the SBR of 2f is much larger than that of 1f. For 3f, the background is comparable to 2f, but the signal is smaller, resulting in reduced SBR, as shown in Fig. 4. It is interesting to notice that the SBR of the nonlinear components remains at least few hundreds throughout the 200 μm imaging depth, manifesting that the imaging depth of plasmonic SAX is only limited by the objective working distance under the examined condition. Therefore, we expect that similar three-fold resolution enhancement continues toward much deeper tissues.
The underlying mechanism of deep-tissue resolution enhancement of SAX is due to the fact that temporal modulation is less affected by tissue scattering/aberration, when compared to spatially structured light (such as donut in STED and spatial modulation in SIM). Having said this, the temporal modulation of SAX may still be degraded by tissue inhomogeneities and in turn limit the maximal depth of maintaining three-fold resolution enhancement. In addition, the optical sectioning capability of the confocal scheme may also decay in deep tissues. From our imaging results, even if the GNSs are dispersed at different depths up to 200 μm in the tissue, the confocal SAX setup is still capable to distinguish signals from different layers since SAX enhances spatial resolution in all three dimensions.17,18 Besides, from the results in Figs. 3 and 4, we did not find a sign of loss in resolution enhancement and contrast at 200 μm depth. Further studies will be necessary to determine the practical limit of plasmonic SAX imaging in observation at deeper positions.
Our previous study suggests that harmonic signals higher than 3f are extractable in SAX microscopy to further improve the image resolution, but with the expense of higher excitation power.17 In this study, we examine the attainable resolution enhancement with harmonic signals as high as 10f for bare GNSs in oil without any tissue. Figure 5 demonstrates representative comparison of GNS images with the fundamental and 10th harmonic signals in the upper row and the corresponding PSF line profiles of a selected particle in the lower row. Remarkably, almost 10-fold resolution enhancement can be achieved with the 10f harmonic signal, manifesting the potential of plasmonic SAX for exceptionally high-resolution imaging. However, in the experiment, at least 5 mW excitation power is required to extract the 10th harmonic signal, while only 0.5 mW is used for the 3rd harmonic signal. The power levels are relatively high compared to typical confocal microscopy. For imaging deeper regions, the excitation power should increase, but the power density to induce saturation at the nanoparticle can be similar to the case of imaging the surface. For a very thick tissue, the excess power may cause photodamage. Nevertheless, no photodamage was observed in our measurement when imaging through 200 μm in depth. This result demonstrates the potential of SAX microscopy to achieve not only further improvement of the spatial resolution, but for live tissue observations.
In terms of tissue imaging, it is more interesting to observe tissue structures, instead of only nanoparticles. There are at least two possibilities to achieve super-resolution imaging on deep tissues itself with the aid of plasmonic nanoparticles. The first one is to label the tissue structures with plasmonic nanoparticles, which has been a well-known approach in the field of electron microscopy.32 The other possibility of imaging tissues with SAX is to use optical effects of tissues themselves, such as harmonic generation and Raman scattering, enhanced by plasmonic resonance. This is indeed an interesting approach and these effects may also show saturation. Since our main scope of this research is to image nanoparticles in tissues, we would like to leave this approach to our next research.
Another concern toward tissue imaging is the potential thermal effect of the nanoparticles. As we have reported in Refs. 22 and 24, local heating on the order of tens of degrees temperature rise is expected on these nonlinear plasmonic nanoparticles. The temperature rise may cause refractive index change of the surrounding medium, inducing additional aberrations, and in turn reduce the image resolution. In addition, high temperature of nanoparticles may cause sample damage. Nevertheless, in our experiments, we did not observe any structural variation during plasmonic SAX acquisition. Since the heat capacitance of nanoparticle is relatively smaller than that of the tissue environment, the real temperature rise in the tissue may be much smaller than the nanoparticle itself.
In summary, we have demonstrated, for the first time, three-fold resolution enhancement inside a biological scattering tissue throughout a depth of 200 μm by plasmonic SAX microscopy. The contrast also maintains throughout the whole imaging depth, suggesting the possibility of extending this method to the observation of thicker tissues. Boosted with the improved resolution in thick tissues and the strong reduction of background signals, plasmonic SAX microscopy has huge potential to be amalgamated with other scanning-based 3D imaging techniques, such as confocal and multiphoton laser scanning microscopy, light sheet microscopy, etc. We anticipate that our study will not only be applicable to biomedical purposes such as tumor imaging during photothermal therapy but also stimulate novel directions in deep-tissue resolution enhancement technologies.
This study was supported by the Ministry of Science and Technology of Taiwan, under Grant Nos. MOST-106-2321-B-002-020 and MOST-105-2628-M-002-010-MY4. This work was supported by JSPS Core-to-Core Program, A. Advanced Research Networks. Shi-Wei Chu acknowledges the generous support from the Foundation for the Advancement of Outstanding Scholarship.