Work

Author response: Intravital deep-tumor single-beam 3-photon, 4-photon, and harmonic microscopy

Year: 2021

Type: peer-review

Abstract: Article Figures and data Abstract Editor's evaluation Introduction Results and discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Three-photon excitation has recently been demonstrated as an effective method to perform intravital microscopy in deep, previously inaccessible regions of the mouse brain. The applicability of 3-photon excitation for deep imaging of other, more heterogeneous tissue types has been much less explored. In this work, we analyze the benefit of high-pulse-energy 1 MHz pulse-repetition-rate infrared excitation near 1300 and 1700 nm for in-depth imaging of tumorous and bone tissue. We show that this excitation regime provides a more than 2-fold increased imaging depth in tumor and bone tissue compared to the illumination conditions commonly used in 2-photon excitation, due to improved excitation confinement and reduced scattering. We also show that simultaneous 3- and 4-photon processes can be effectively induced with a single laser line, enabling the combined detection of blue to far-red fluorescence together with second and third harmonic generation without chromatic aberration, at excitation intensities compatible with live tissue imaging. Finally, we analyze photoperturbation thresholds in this excitation regime and derive setpoints for safe cell imaging. Together, these results indicate that infrared high-pulse-energy low-repetition-rate excitation opens novel perspectives for intravital deep-tissue microscopy of multiple parameters in strongly scattering tissues and organs. Editor's evaluation Nonlinear microscopy is in the unique position that high-resolution images of cells and other tissue components can be obtained in live tissue. However, scattering and absorption limit the penetration depth and thus the level of 3D information that can be achieved. In this manuscript, the authors develop a new approach that can accomplish deeper imaging in complex specimens, which has potential to provide new insight into biological mechanisms within living tissues and whole organisms. https://doi.org/10.7554/eLife.63776.sa0 Decision letter eLife's review process Introduction Intravital microscopy enables studies of the physiology and malfunction of live cells in multicellular organisms (Ritsma et al., 2012). Several excitation and detection modalities have been applied for multiparameter three-dimensional (3D) imaging in live tissues, yielding different compromises between imaging depth and resolution. Linear imaging modalities such as confocal microscopy enable a penetration depth less than 100 µm (Helmchen and Denk, 2005), which restricts their application to low-scattering, thin or surface-specific tissue models, such as the mouse inner ear and chick embryo (Deryugina, 2016; Tomo et al., 2007; Figure 1a, Supplementary file 1a). Taking advantage of reduced scattering and absorption of the excitation light and selective excitation of the focal plane (Helmchen and Denk, 2005; Diaspro et al., 2005), 2-photon (2P) microscopy based on high-frequency near-infrared excitation pulses (80 MHz, ~100–200 fs, 800–1300 nm, further referred to as lowIR) provides 100–400 µm imaging depth in highly scattering tissues, for example, liver (Ritsma et al., 2012) and skin (Andresen et al., 2009), and up to 500–800 µm in less scattering tissues, such as mouse brain cortex (Helmchen and Denk, 2005). Figure 1 with 5 supplements see all Download asset Open asset Multimodal microscopy of fluorescent skin tumor xenografts in vivo showing up to four fluorescent and two label-free tissue morphology channels, simultaneously excited with a single highIR wavelength. (a) Jablonski diagrams (left) and scaled excitation/emission spectra (right) of 1-, 2-, 3-, and 4P-excited fluorescence and higher harmonic generation. Diagram 1–3 (left to right): 1- up to 3P excitation of eGFP with 488 nm continuous, 910 nm lowIR, and 1300 nm highIR, respectively. Diagram 4–7: excitation with 1650 nm results in signals of a distinct emission wavelength, including 4P-excited eGFP. S0 and S1, ground level and first-excited electron state, respectively, with vibrational levels (thin lines). Dashed lines, virtual states. Thin–thick upward arrows: low–high photon density required for excitation. Downward arrows: emission. (b) Five-channel 1300 nm excited (left) and six-channel 1650 nm excited (middle and right) images were taken in the center of fluorescent tumors through a dermis imaging window. Images were selected from median-filtered (one-pixel) z-stacks. The excitation wavelength, calculated pulse energy at the sample surface, and imaging depth z are indicated (top). Cell nuclei containing a mixture of mCherry and Hoechst appear as green (right). Details from images left to right: day-10, day-13, and day-14 tumor; 9.2, 4.5, and 4.4 s image acquisition time; 24, 12, and 12 µs pixel integration time; 0.36, 0.46, and 0.46 µm pixel size. (c) Zoomed xy-plane with individual channels (left) and orthogonal (xz-) projection (right) from (b), dotted rectangle. (d) Relation between measured emissions and excitation energy reveals the order of the excitation processes. Normalized emission intensity (S) as a function of pulse excitation energy (E), for TagRFP, SHG, Hoechst, third harmonic generation (THG), and eGFP recorded with an excitation wavelength of 1650 nm. Data was fitted with S(E) = A⋅En, with A the proportional factor and n the order of the excitation process (indicated as numbers in the figure). For curve fitting, excitation intensities at the sample surface below the threshold of physical damage (14 nJ) for SHG and THG and up to their saturation limit (7.6 nJ, eGFP; 6.9 nJ, TagRFP and Hoechst) for fluorophores were used. Images were acquired at the same position as panel (a, right), except for the fit line of eGFP, which was retrieved from a different dataset (Figure 1—figure supplement 4). Bars, 25 µm (b); 12.5 µm (c). Source data files: Figure 1—source data 1 and Figure 1-source data 2. Figure 1—source data 1 Numerical data, calculations and templates used to generate graph panels. https://cdn.elifesciences.org/articles/63776/elife-63776-fig1-data1-v1.zip Download elife-63776-fig1-data1-v1.zip With increasing imaging depth in tissue, 2P imaging requires to increase the excitation power in order to compensate for tissue-induced light scattering, absorption and aberrations, to maintain a constant intensity at the plane of focus. However, the required exponential increase in excitation power may provoke light-induced tissue damage (Débarre et al., 2014). Heating caused by linear absorption can be minimized by redistributing the excitation power into a reduced number of pulses per second and a higher energy per pulse (further referred to as highIR), which enhances the excitation efficiency per pulse. For example, reducing the pulse repetition rate from 80 to 1 MHz while increasing the pulse energy is an effective means to reach larger depths (Beaurepaire et al., 2001; Theer et al., 2003), albeit at the cost of reduced imaging speed, because 80 times less excitation events per second are introduced into the sample. Ultimately, 2P imaging depth in densely labeled fluorescent samples is limited by out-of-focus background produced in superficial regions, which at large depths becomes dominant over the signal produced in the focus (Helmchen and Denk, 2005; Theer and Denk, 2006). To reduce these confounding effects, shifting the excitation wavelength to the infrared optical windows at ~1300 or 1650–1700 nm was shown to reduce light scattering and, by 3-photon (3P) excitation, improve confinement of the scan focus and thereby reduce out-of-focus background signal (Xu et al., 1996; Hell et al., 1996). This approach is compatible with commonly used fluorophores (Deng et al., 2019; Hontani et al., 2021), and was demonstrated to increase penetration in densely labeled brain tissue by 50–100%, reaching imaging depths beyond 1 mm in the brain of live mice (Horton et al., 2013; Ouzounov et al., 2017; Yildirim et al., 2019; Weisenburger et al., 2019). In fundamental and preclinical cancer research, intravital microscopy is an important strategy to record dynamic processes of the tumor microenvironment, including tumor cell invasion, metastasis, and interactions with the tumor stroma. Intravital microscopy of the tumor microenvironment faces several tissue-intrinsic challenges. Tumors are dense, opaque, three-dimensional objects with a size of millimeters to centimeters, which causes strong light scattering (Andresen et al., 2009). In addition, analysis of the tumor microenvironment requires to detect multiple cell types, their morphology and activation states simultaneously, and tissue structures over time through a body window (Weigelin et al., 2021). HighIR combined with 3P microscopy enables the label-free detection of metabolites, tissue morphology, and leukocyte migration in mammary tumor microenvironment in vivo (You et al., 2018b). However, the relevance of 3P intravital microscopy for multiparametric detection of tumors expressing fluorescent reporters, as used in preclinical cancer research, and detection of tumor cells deeply inside the tumor has not been explored. We here apply 1300 and 1650–1700 nm highIR excitation to achieve 3- and 4P and higher harmonics microscopy of live fluorescent tumors in mice and other strongly scattering tissues, such as bone. We report improved multiparameter detection of cells and complex tissues, a 2-fold gain of penetration depth at non-toxic laser intensities, and established criteria to minimize photodamage that enable long-term time-lapse imaging of live-cell biology. Results and discussion 3P excitation setup characterization In contrast to the well-established 2P microscopy lowIR excitation regime (100–200 fs pulses with sub-nJ energy density at the focus), efficient 3P microscopy requires shorter (sub-100 fs) pulses at the focus with ~5–10 times higher energy (König et al., 1999) (highIR excitation) to generate comparable excitation of fluorophores (Xu et al., 1996; Wang and Xu, 2020; Supplementary file 1a, Figure 1a). To keep the average power and associated sample heating at an acceptable level, 3P microscopy applies lower pulse repetition rates (1–2 MHz) compared to pulse repetition rates commonly used for 2P microscopy (80 MHz) (Beaurepaire et al., 2001; Theer et al., 2003). The source used in this study consisted of a fiber laser pumping an optical parametric amplifier (OPA) providing 1 MHz, sub-100 fs pulses at 1300 or 1650 nm wavelengths. Excitation was focused with a high-NA objective, resulting in a resolution better than 0.8 µm in the lateral and 3 µm in the axial dimension (Figure 1—figure supplement 1). Physiological saline was used as immersion medium for high compatibility with the tissue and observation window. Since part of the excitation pulse energy (Eobj) was absorbed by the immersion medium, we calculated the pulse energy at the sample surface (E) as a reference for each experiment (Figure 1—figure supplement 2). We compared highIR excitation at 1300 and 1650 nm with lowIR excitation in the range 1100–1300 nm, the current standard for in-depth multiphoton imaging (Andresen et al., 2009). Using the mouse brain as a reference (Horton et al., 2013; Weisenburger et al., 2019), we confirmed that 1650 nm highIR excitation enables to resolve structural details with third harmonic generation (THG) contrast at depths encompassing the entire cortex (>1 mm), which outperforms the imaging depth obtained with 1270 nm lowIR excitation by almost a factor of two (Figure 1—figure supplement 3b,c). Since both high- and lowIR excitation measurements were using a 3P process to generate contrast, the superior performance with highIR excitation can be attributed to the enhanced excitation efficiency rather than to reduced out-of-focus background. We calculated the effective attenuation length (le), which is the mean distance travelled by light before being scattered or absorbed by the tissue (Wang et al., 2018b). In the neocortex, le was 336 µm for 1650 nm excitation and 280 µm for 1280 nm excitation (Figure 1—figure supplement 3d), in agreement with previous work (Horton et al., 2013; Wang et al., 2018b). Simultaneous multicolor 3P, 4P, and harmonic generation microscopy with a single excitation beam When applied to multicolor-labeled HT-1080 sarcoma tumors in the deep dermis, excitation with 1300 or 1650 nm highIR pulses produced simultaneous distinct signals from fluorescent proteins (eGFP, TagRFP, mCherry), a far-red vascular tracer (AlexaFluor680-conjugated 70-kD-dextran) and blue fluorescence (Hoechst 33342), together with second harmonic generation (SHG) and THG signals (Figure 1b–c, Figure 1—video 1). To understand the nonlinear processes underlying this unusually broad spectrum of fluorophores excited with a single laser line, we investigated the order of the excitation process for each signal (Figure 1a). We recorded the emission intensity in response to increasing excitation energy E and estimated the order of dependence between emission intensity and E by fitting the data to a power law (Figure 1d, Figure 1—figure supplement 4 and Supplementary file 1b). As internal controls, SHG and THG signals were used as reference for second- and third-order processes (Cheng et al., 2014). When excited at 1650 nm, the red fluorophores TagRFP, mCherry, and AF680 followed a cubic dependence on excitation power, while green and blue fluorophores eGFP and Hoechst followed a quartic dependence, consistent with respective third- and fourth-order processes (Cheng et al., 2014). This shows that 3- and 4-photon-excited fluorescence (3PEF, 4PEF) along with multi-harmonic SHG and THG signals were obtained simultaneously using 1650 nm excitation, resulting in six-channel images in a single scan (Figure 1b). Although independent control of each respective signal strength is not possible, a key advantage of this one-shot excitation scheme is that it is free of chromatic aberration, a common problem in multi-beam imaging (Theer et al., 2014). Characterization of phototoxicity and bleaching The high pulse energy required for efficient highIR excitation is a double-edged sword. Although it enables multi-modal excitation as seen above, pulsed light can induce nonlinear and linear perturbations. First, the energy density within the beam focus, which causes nonlinear interactions between a strong electric field and matter, may induce physical damage by ionization (Konig et al., 1997), toxic reactive oxygen species, and photobleaching (Débarre et al., 2014; Tirlapur et al., 2001). This focused energy may further cause local transient heating of the focal volume during the excitation pulse, caused by water absorption (Qiu et al., 2017). This type of phototoxicity is dependent on the pulse energy (nJ) at the focal plane. Second, phototoxicity may arise from linear absorption of light above and below the focal plane, which may cause accumulation of heat in a larger tissue volume (Débarre et al., 2014; Podgorski and Ranganathan, 2016; Wang et al., 2018a; Rowlands et al., 2017). Sample heating is proportional to the average power of the incident excitation beam. Previous studies have investigated phototoxicity mechanisms with 1300 nm highIR excitation in the brain, and have established safety thresholds for 3P imaging of neuronal activity in this wavelength range (Ouzounov et al., 2017; Yildirim et al., 2019; Wang et al., 2020). In this work, we investigated the mechanisms and onset of photoperturbation with 1700 nm highIR excitation in tumor models in vitro (tumoroids), and with 1300 and 1700 nm highIR excitation in mouse dermis tumors in vivo and achieved safe live-cell imaging conditions. As a readout for cell stress, we continuously recorded the intracellular Ca2+ responses of live tumor cells expressing a Ca2+ sensor GCaMP6 (Chen et al., 2013) in combination with the cell-death indicator Sytox Green (Figure 2a). GCaMP6 fluorescence was monitored using lowIR 2P excitation (920 nm, 80 MHz) in parallel to 1700 nm highIR exposure (see Materials and methods). Intracellular Ca2+ responses represent a sensitive readout for non-lethal cell stress, caused by opening of calcium channels as well as by transient membrane damage caused by reactive oxygen species and other cytotoxic events (Hopt and Neher, 2001; Iwanaga et al., 2006). In a collagen-based 3D phantom containing a reconstituted cell-rich tumoroid, spontaneous reversible Ca2+ fluctuations and few Sytox-positive damaged cells were detected before highIR exposure (Figure 2b, frame #70). As a positive control, intermittent highIR scanning at toxic intensity (17 nJ) caused rapid-onset Ca2+ elevation in cell subsets followed by gradual increase of the Sytox signal (Figure 2b, frames #102–201). High exposure was further confirmed by significant photobleaching of mCherry emission from cell nuclei (Figure 2b, arrowheads). After termination of toxic highIR exposure, the Ca2+ elevation persisted, and an increasing number of cells became Sytox-positive (Figure 2b, frame #425). To study the toxicity of highIR exposure in detail, Sytox-positive and -negative cells at the endpoint (frame #425) were grouped and the kinetics of Ca2+ and Sytox responses of both subsets were quantified (Figure 2c and Figure 2—figure supplement 1a). The Sytox-negative, surviving cells showed a transient Ca2+ elevation which declined to half of the maximum amplitude after cessation of highIR exposure, whereas Sytox-positive cells showed an accelerated and transient Ca2+ elevation upon onset of highIR exposure in combination with delayed nuclear Sytox increment (Figure 2c). In Sytox-positive cells, the GCaMP6 signal level eventually declined to its initial level before highIR exposure, likely as a consequence of persisting membrane damage with leakage of GCaMP6 in parallel to Sytox entry. At the single-cell level, Ca2+ elevation was nearly instantaneous with the onset of laser exposure in Sytox-positive cells, followed by Sytox elevation thereafter with a large spread of the onset time (Figure 2—figure supplement 1b). Thus, the Ca2+ signal reliably indicated phototoxic stress, prior to detection of cell death by Sytox and even in cells which remained viable at the experimental endpoint. Figure 2 with 4 supplements see all Download asset Open asset Phototoxicity and interference with biological function induced by continuous scanning with highIR excitation. (a) Experimental approach for cell viability detection in vitro. B16F10/GCaMP6/H2B.mCherry (Chen et al., 2013) melanoma tumoroid embedded in 3D collagen was monitored through a coverglass. Sytox Green was supplemented to the medium. Both GCaMP and Sytox signals were continuously monitored (3.3 s/frame, 6 µs pixel integration time) at ~40 µm penetration depth using low pulse-energy excitation (920 nm). During part of the measurement, the sample was simultaneously exposed to 1700 nm highIR excitation, approximately at the same focal plane (I). ( b) Representative frames from continuous scanning with the indicated excitation pulse energy and detection of emitted signals. Details from dotted squares: (1) spontaneous reversible Ca2+ activity (green); (2) highIR-excited fading mCherry signal from the nuclei (red); (3) strong highIR-induced Ca2+ response in the majority of the cells and bleached, decreasing mCherry signal; and (4) Sytox-positive nucleus (yellow). (c) Intensity curves of Sytox left and GCaMP signals in cells Sytox-positive nuclei or not at the endpoint by (frame #425). Data show the mean intensity of single-cell For see Figure 2—figure supplement and (d) power of GCaMP for The highIR-induced of cells was quantified over time for cells from n = 4 and cells from n = 4 conditions. = Data indicate the and of GCaMP and Sytox to excitation plane (I). A of 80 µm depth µm was taken after the endpoint of highIR exposure shown in at the highIR exposure plane at the line in the in excitation plane of phototoxicity of highIR in tumor after 6 or of GCaMP signal was continuously monitored s/frame, 6 µs pixel integration time) in an µm deep focal plane using lowIR excitation nm). The sample was simultaneously exposed to highIR excitation at the same focal plane. Ca2+ in individual Representative frames from the 1300 nm highIR exposure at pulse at the sample surface below or above nJ) the toxicity reversible Ca2+ signal in single as seen in of cells in the field of Ca2+ signal at frame as seen in of multiple cells increasing Ca2+ in of The frame number is Ca2+ signal as a function of time for increasing highIR pulse at the sample surface, recorded with an excitation wavelength of 1300 nm or 1650 nm and and gradual Ca2+ respectively, to cell as in signal was retrieved after over the image background and to the highIR-excited frame 1300 nm and 1650 nm, dotted lines). The number of cells per field is µm Source data files: Figure data 1 and Figure data 2. Figure 1 of data into Download Figure 2 microscopy image and analysis Download Figure 3 cell upon highIR exposure as a function of Download Figure 4 cell upon highIR exposure as a function of imaging Download Figure data 1 Numerical data, calculations and templates used to generate graph panels. Download Using both GCaMP6 and Sytox we to the highIR conditions by live tumor cells upon using a highIR energy approach followed by of the highIR-induced Ca2+ cell for each Materials and methods for further For highIR pulse from and nJ, the of cells with Ca2+ was whereas at 12 this was increased (Figure Thus, a threshold for excitation energy in the focus of up to at the sample surface for safe time-lapse imaging inside the tissue, at µm deep inside the sample. To for linear phototoxicity of the highIR focal plane, depth were obtained before and after time-lapse measurements at highIR exposure Both Ca2+ and Sytox signals were within the highIR excitation plane, without in regions above and below (Figure and Figure 2—figure supplement This shows that continuous scanning with 1700 nm highIR excitation power of (17 1 MHz) at the sample surface can be used without linear To the for phototoxicity for intravital we exposed mouse dermis tumors to 1300 and 1650 nm highIR excitation and monitored their Ca2+ responses (Figure During continuous highIR exposure for at the sample surface below or to or the Ca2+ signal background with spontaneous Ca2+ fluctuations (Figure Figure 2). excitation for 1300 nm, 12 for 1650 induced Ca2+ elevation in cell subsets (Figure and Figure arrowheads). Ca2+ responses from the background fluctuations by their or gradual increase of signal (Figure and arrowheads). at exposure above the Ca2+ signal the onset of (Figure and Figure 1 Figure or (Figure 2—figure supplement 2a). to damage induced by we applied average power levels at the sample surface of at times 1 MHz) live-cell and in vivo which in the brain to limit tissue heating below and Ranganathan, 2016; Rowlands et al., 2017). Thus, we established a limit for power to be used for excitation in tumors to below phototoxicity levels and showed that higher induce different of damage et al., et al., To 3D scanning was over 3D using highIR exposure within the toxic nJ, 1300 12 nJ, 1650 or the toxic eGFP intensity declined by after 25 and mCherry intensity declined by after two (Figure 2—figure supplement the non-toxic mCherry intensity by after while eGFP signal remained This level of photobleaching may limit applicability for imaging with frame as required for Ca2+ in the brain et al., 2019), acceptable for with lower frame in the up to and for which are imaging for tumor cell kinetics and response in tumor models et al., 2020; et al., 2020). multiparameter microscopy with highIR excitation We investigated highIR excitation at 1300 and 1650 nm provides an advantage for imaging regions, with to lowIR excitation at nm using a parametric combination (Figure To achieve constant emission with increasing imaging we increased the excitation energy E using eGFP 1300 and TagRFP and 1650 as a while below the phototoxicity thresholds as above (Figure 3P and 4P highIR excitation of eGFP and TagRFP labeled cells at depths beyond as compared to µm for 2P lowIR excitation (Figure were reliably obtained at a depth of µm inside the tumor using 1300 and 1650 nm highIR excitation, not with the nm lowIR source (Figure To the imaging depth limit for the different we quantified the of fluorescent as a function of imaging the remained above the detection limit a depth of µm using the 2P lowIR excitation the 3P highIR excitation regime to reach depths of µm or µm (Figure to a nearly 2-fold increase as compared to The of deep-tissue microscopy also on which is by scattering and of the incident excitation beam (Cheng et al., 2020). We thus the axial resolution on imaging depth and excitation For and resolution remained higher with increasing imaging depth as compared to the resolution achieved by which declined beyond µm (Figure Thus, and modalities provide improved resolution in strongly scattering tissues compared to which may be by the superior confinement of the 3P or 4P excitation processes (Xu et al., 1996; Hell et al., 1996). In addition, light scattering and may the of the focus and, resolution in a (Cheng et al., 2020). To the attenuation of with increasing tissue penetration in we measured the fluorescence intensity as a function of increasing imaging depth and the effective attenuation length le (Figure le remained constant over of that the tumor was over this depth When the excitation wavelength from to 1300 or 1650 nm, le increased from to or

Source: eLife

DOI: 10.7554/elife.63776.sa2