2.1.

Adaptive Harmonic Generation Microscope

Several factors must be considered in achieving the best HG images of photosensitive specimens. One of the important aspects is the selection of excitation wavelength. We employed a Cr: Forsterite laser system (CrF-65P, Avesta Project, Moscow), which emits 65-fs pulses at a repetition rate of 76 MHz. The wavelength was centered at 1230 nm, and the output average power was ∼200 mW. The choice of the illumination laser wavelength (1230 nm) was influenced by the relatively low absorption of this wavelength by biological tissues and the convenient placement of the SHG and THG around 600 and 400 nm, respectively, permitting the use of standard optics and detectors. The following factors all lead to aberrations and, thereby, reduction in the focal spot intensity: The use of microscope objectives designed for visible wavelengths, refractive index variations between the coverglass and the immersion medium, and keeping the objective lens at a higher temperature for the embryo culture.¹⁸ Therefore, we incorporated a deformable mirror (DM) in the microscope, to compensate for such system-induced optical aberrations.

Figure 1a depicts an outline of the experimental setup. The diameter of the beam at the laser output was 1.2 mm and was expanded 4× using a telescope arrangement. The expanded laser beam was scanned by a pair of galvanometric mirrors (VM1000, GSI Lumonics, Bedford, Massachusetts), which were imaged onto the DM (MIRAO 52-e, Imagine Eyes, Orsay, France) by a 2.5× telescope. The entire active area of the DM was then again imaged onto the pupil plane of the microscope objective using a pair of telescope arrangements with total magnification 0.75. This imaging was critical for the calibration of the DM and for the accurate generation of wavefront aberration modes at the pupil plane. The shape of the DM can be controlled by actuators located beneath its silver-coated membrane. All the lenses selected for the telescopes were near-infrared (NIR)–coated doublets to maximize the transitivity of the laser beam and to minimize the aberrations. The expansion of the beam was designed in such a way as to achieve a truncated Gaussian illumination profile at the back aperture of the objective lens (the intensity variation across the pupil plane was no more than 30%). This was chosen to provide a compromise between optimum resolution, which would be provided by an overexpanded beam, and higher focal intensity, corresponding to an underfilled aperture. We selected an Olympus UApo/340 water immersion objective [40×, numerical aperture (NA) = 1.15] for focusing the illumination light on the specimen. Although this objective was optimized for UV wavelengths, its transmission at the 1230-nm laser wavelength (nearly 40%) was higher than a range of other standard microscope objectives that were tested. The average laser power available at the sample plane was ∼20% of the laser output power.

Fig. 1

(a) Schematic of the adaptive harmonic generation microscope. Lx, lens; Mx, mirror; BSx, beam splitter; G, Galvo mirrors; DM, deformable mirror; O, objective; S, specimen; C, condenser; Dx, dichroic; PMTx, photomultiplier tubes; FM, flip-mirror. He–Ne laser (dashed outline) is used for DM characterization, and this path is disabled during imaging. Dotted lines show the LED illumination path for wide-field transmitted light imaging. The incubator maintains a stable temperature at 37 °C around the sample stage. (b) A magnified view of the region around the sample. A small plastic chamber around the culture dish, into which a humidified mixture of 5% [TeX:] \documentclass[12pt]{minimal}\begin{document}$\rm { CO_{2}}$\end{document} ${\mathrm{CO}}_2$ in air is supplied to maintain the pH of the culture medium. (c) Illustration of the culture dish showing the placement of the embryos.

The HG signal emission pattern is sensitive to the relative orientation of the features with in the specimen with respect to the focal spot.¹¹ Therefore, the harmonic light was collected in a transconfiguration by a high NA, oil immersion condenser (NA = 1.4, Olympus, Tokyo, Japan). A dichroic beam splitter (FF495-Di02-25×36, Semrock, USA) was used to transmit the SHG wavelength while reflecting the THG signals. The HG signals were then recorded on two separate photoncounting photomultiplier tubes (PMTs). A bandpass filter FF01-625/26 (Semrock) was inserted before the SHG PMT (H7422P-40 Hamamatsu, Japan) and another bandpass filter FF01-417/60 was placed before the THG PMT(P30PC-01, SensTech, Berkshire, UK) to further improve the signal to noise ratio.

Scanning in the axial (z) direction was enabled by a piezoactuator attached to the sample stage. A software-controlled shutter was inserted in front of the laser to block the laser beam while not scanning. DM control, 3-D, and time-lapse data acquisition and shutter control were performed with LabView software. The DM calibration was performed as described in Ref. 19. A helium-neon laser beam (633 nm) was used for this purpose. This path was disabled during HG imaging by removing the beam-splitter cubes. A green LED source and a CCD camera were employed for wide-field transmitted light imaging of the specimen for sample viewing and alignment.

2.2.

Mouse Embryo Preparation and Culture

Early mouse embryos are extremely sensitive to the their culture environment. Not only do they have strict nutritional requirements, they also require constant humidity to help minimize evaporation of the culture medium. Because they normally develop in a controlled uterine environment, they do not tolerate changes in temperature and pH well. Embryos are normally cultured in tissue-culture incubators, which have a relatively large volume maintained at 37 °C, 100% humidity and an atmosphere of 5% [TeX:] \documentclass[12pt]{minimal}\begin{document}$\rm { CO_{2}}$\end{document} ${\mathrm{CO}}_2$ in air. This large volume buffers any fluctuations in environmental conditions.

In order to support embryo growth under our microscope, we constructed a Plexiglas enclosure around the microscope to maintain the temperature. A fan heater (FLH 400 W, Pfannenberg, Hamburg, Germany) was mounted inside the enclosure and connected to a custom-built temperature-controller unit to maintain a stable temperature of 37 °C around the sample stage. Vibration-dampening air cushions were inserted on the heater mount as well as on the base of the enclosure to remove the vibrations due to the fan. In order to create a microenvironment with the correct atmospheric gas mix, the condenser and sample were enclosed by a small plastic chamber into which a humidified mixture of 5% [TeX:] \documentclass[12pt]{minimal}\begin{document}$\rm { CO_{2}}$\end{document} ${\mathrm{CO}}_2$ in air was supplied to maintain the pH of the culture medium [Fig. 1(b)].

Embryos in a tissue-culture incubator are normally cultured in a drop of medium covered with mineral oil to prevent evaporation. The mineral oil allows gas exchange; thus, the medium remains at the correct pH in the course of culture. To culture embryos in our microscope, we had to design an approach that accommodated both culturing and imaging needs. The medium for embryo imaging was KSOM (MR-050P-5F, Millipore, Billerica, Massachusetts) supplemented with essential amino acids (11130-036, Invitrogen, Paisley, UK), nonessential amino acids (Invitrogen 11140-350), and sodium pyruvate. Embryos were placed in a drop of this medium in a glass-bottom dish (Mattek, Ashland, Massachusetts). Because harmonic generation is primarily forward propagated, the drop of medium was covered with a coverglass. To provide space for the embryo, the coverglass was supported by mylar spacers of 200-μm thickness that were affixed using nontoxic silicone grease. The dish was then filled with embryo-grade mineral oil (M5310, Sigma St. Louis, Missouri) to prevent evaporation of the medium [Fig. 1(c)]. The coverglass over the drop of medium had to be made small enough (approximately 25×25 [TeX:] \documentclass[12pt]{minimal}\begin{document}$\rm {mm^{2}}$\end{document} ${\mathrm{mm}}^2$ ) to permit efficient gas exchange between the medium and the 5% [TeX:] \documentclass[12pt]{minimal}\begin{document}$\rm { CO_{2}}$\end{document} ${\mathrm{CO}}_2$ environment. Despite not having the optimum refractive index, the mineral oil used to cover the culture medium was also used as the immersion medium for the condenser lens, because standard immersion oils were found to be toxic to the embryos.

2.3.

Aberration Correction

We used the THG signal generated from coverglass–culture medium interface to compensate for the system-induced aberrations in the HG microscope. The aberration correction was performed in a sequential manner, based on modal wave-front sensing as described in Ref. 16. This correction procedure in total, took nearly 1 min. This consists predominantly of spherical aberration, partially due to incomplete coverglass thickness compensation in the objective lens and partially due to the use of this objective at an out-of-specification wavelength. Compensation of system aberrations was performed once before the imaging of each specimen, and the same correction was used throughout the imaging process. Further correction of specimen-induced aberrations was not performed during this study. Compensating system aberrations not only improve the image resolution but also resulted in overall signal improvement of nearly 40%. This allowed us to reduce the laser average power at the sample down to nearly 35 mW without compromising image brightness. This is considerably lower than the power levels used in the previously reported THG experiments.^{14, 15}

2.4.

Imaging Properties of the Microscope

The design of the microscope required various pragmatic compromises between optical and biological requirements. The optical components were not necessarily optimized for the required imaging conditions, and the ability to simultaneously image several embryos demanded a wide field of view. We characterized the imaging properties in terms of field of view, uniformity of illumination, field curvature, and resolution.

THG signal was obtained over a field of view of diameter 550 μm. This allows simultaneous observation of several early-stage embryos. We observed that the microscope had a curvature of field as demonstrated in Figs. 2a and 2b, which show THG signal obtained from a coverslip–air interface when placed perpendicular to the beam-propagation direction. Figure 2a is an optical section from a stack of images, and Fig. 2b is an XZ view along Y = 0. The maximum distortion of the field was 15.5 μm over the 550-μm width. We attribute this effect to the use of the objective lens outside of its design specifications. Although the curvature is large across the whole field, it is far less significant over the ∼100 μm width of each embryo specimen. As such, we did not consider this to have a detrimental effect on our results. If necessary, the measured focal plane distortion could be incorporated into future morphological analysis. THG signal intensity drops from the center to the periphery of the field, as shown in Fig. 2c. We attribute this effect to the field-dependent aberrations as well as to the vignetting at extreme field positions. The aberration correction was performed only over a region of nearly 50×50 [TeX:] \documentclass[12pt]{minimal}\begin{document}$\rm {\mu m^{2}}$\end{document} $\mu {\mathrm{m}}^2$ around the center of the field of view. The resulting correction phase amplitude had a root-mean-square (rms) value of 0.74 rad. After aberration correction, the axial THG intensity profile measured along the center of the field of view had a FWHM of 1.34 μm [Fig. 2(d)]. This is comparable to the corresponding theoretical value 1.15 μm for an unaberrated system, with uniform pupil illumination. The small discrepancy could be due the use of truncated Gaussian illumination in the experiment.

Fig. 2

(a) THG signal from a cover slip–air interface placed perpendicular to the laser beam direction and (b) the corresponding XZ view demonstrating the curvature of field of the microscope. (c) THG signal variation across the entire field of view. (d) The axial THG intensity profile along the cover glass–air interface of a region near the center of the field of view has FWHM of 1.34 μm. Scale bar in (a) is 100 μm.

3.1.

HG Characterization of Early Embryonic Development

Bacause embryos undergo dramatic changes in the course of early development, we characterized HG signals generated by live embryos at various stages; 0.5-day-old zygote, 1.5-day-old two-cell embryo, 2.5-day-old uncompacted morula, 3.5-day-old blastocyst, 4.5-day-old peri-implantation stage embryo, and 5.5-day-old postimplantation embryo. Embryos at these stages were found to be transparent to the 1230-nm wavelength light, and therefore, it was possible to observe the entire embryo. Optical sections were taken typically at a resolution 0.2–0.3 μm/pixel and were acquired at a rate of approximately 3–4 s/frame with z step of typically 0.5–1 μm. Three-dimensional rendering of images was done using Volocity software (Improvision, Coventry, UK) and ImageJ.²¹

Figure 3 shows combined SHG (magenta) and THG (gray) images. Figures 3a, 3b, 3c are single optical sections from 0.5- (single cell), 1.5- (2 cells), and 2.5- (8 cells) day-old embryos, respectively, and Figures 3(a^′)–3(c^′) are the corresponding 3-D opacity renderings. (In the 3-D images, the intensity is scaled logarithmically). It is evident that SHG and THG signals provide complementary information. On the basis of their shape, location, and similarity to published reports,²⁰ we believe the SHG signal observed in the early-stage mouse embryos is most likely to be fibers of the central spindle (blue arrows) during the formation of the second polar body (white arrow) and cleavage division of blastomeres.

Fig. 3

HG images of 0.5- (single cell), 1.5- (two cells), and 2.5-(morula) day-old mouse embryos. (a–c) are optical sections and (a c) are the corresponding 3-D opacity rendering (intensity in log scale). The SHG signal [blue arrows: pointing top right in (a^′), middle part in (b^′) and in the lower left part of (c) and (c^′)] is from central spindles, whereas THG is predominantly from lipid droplets [red arrow head in the left side of (a)]. Features like the plasma membrane (yellow arrow head in the right side of (a-c)), nucleus [asterisk in (a)], nucleoli [red arrow pointing left in (a–c), sperm tail (yellow arrow on the top left part of (a^′) and upper right part of (c^′)], and second polar body [white arrow on the upper middle part of (a)] are also demarcated by THG microscopy. The scale bar is 30 μm. (Color online only.)

THG signals were observed from several features in the specimen. Our colocalization experiments have suggested that the majority of THG contrast generated by the preimplantation embryo is from lipid droplets.¹⁷ As can be seen from Fig. 3, the one-cell zygote, two-cell embryo, and eight-cell morula are all roughly similar, densely packed with numerous small lipid droplets. At these stages, the plasma membrane is faintly visible (yellow arrowhead), though it is overshadowed by the much stronger cytoplasmic lipid droplet (red arrowhead) signal. The nuclear membrane, despite being a double membrane, does not generate detectable THG signal, but nuclei are visible as areas of absence of THG signal (yellow asterisk). It is interesting to note that the nuclear membrane in zebrafish embryos do generate THG signal.¹⁴ One possible reason why we do not detect a nuclear membrane signal in early-stage mouse embryos is because of a difference in the optical properties of the nuclear membranes and/or the surrounding regions. Another more likely reason is that the THG signal from the nuclear membrane in early mouse embryos is obscured by the very strong cytoplasmic THG signals. There seems to be a great deal less cytoplasmic THG signal in zebrafish, which might be why nuclear membrane signal is more readily visible in the zebrafish. In further support of this idea, we did detect nuclear membranes in later-stage mouse embryos, which have reduced cytoplasmic signal. The nucleoli with in the nucleus generate THG signal (red arrows) but become progressively more difficult to detect as the embryo develops. Extracellular features, such as the sperm tail (yellow arrow), give a relatively stronger THG signal, and a fainter THG signal is observed from zona-pellucida (magenta arrow).

In Fig. 4, HG images from 3.5-, 4.5-, and 5.5-day-old embryos are shown. Figures 4a, 4b, 4c are single optical sections, and Figs. 4(a^′)–4(c^′) are the corresponding 3-D opacity renderings (in log scale). By day 3.5, the embryos become blastocysts [Fig. 4(a)], characterized by a cavity called blastocoel (yellow asterisk), inner cell mass (yellow arrow), and an outer layer called trophectoderm (red arrow). In blastocysts, lipid droplets are fewer in number, larger, and rounded compared to those in earlier stages and are present both in trophectoderm and inner cell mass. The plasma membrane is also visible with high contrast in blastocysts [see Video 1 showing a 3-D volume rendering (maximum intensity projection) of a blastocyst].

Fig. 4

HG images of 3.5- (blastocyst), 4.5- (peri-implantation), and 5.5- (postimplantation) day-old mouse embryos. (a–c) are optical sections and (a–c) are the corresponding 3D opacity rendering. In (a), the yellow arrow to the right marks the inner cell mass, the red arrow to the left the trophectoderm, and the asterisk the blastocoel cavity (see Video 1 for full 360-degree rotation of a blastocyst). In (b), the red arrow (top middle part) marks a cell nucleus and the white arrow (at the bottom) marks a lipid droplet. In (c), the blue arrow at the right marks HG signal from the basolateral aspect of the visceral endoderm and the asterisk marks the proamniotic cavity (see Video 2 for full 360-degree rotation of a 5.5-day-old embryo). The scale bar is 50 μm. (Color online only.)

Video 1

3-D volume rendering (maximum intensity projection) of a 3.5-day-old (blastocyst) mouse embryo (QuickTime, 2.7 MB)

Video 2

3-D opacity rendered view of a postimplantation stage (5.5-day-old) mouse embryo. (QuickTime, 954 KB)

We found that the THG signal levels from the features such as plasma membrane, nuclear membrane, and lipid droplets vary in a stage-and-tissue-specific manner as the embryo develops. This can be due to the dynamic changes of their optical property differences with the surrounding media. In 4.5-day-old embryos, we start to see HG signal from the nucleus [red arrow in Fig. 4b] and relatively less signal from lipid droplets, which is localized predominantly to the mural trophectoderm [white arrow in Fig. 4b], which mediates invasion of the uterine wall during implantation. We do not know the significance of lipid localization specifically to the mural trophectoderm, but it may possibly have a role in implantation. The developing embryo is marked with a yellow asterisk and shows reduced amount of THG signals.

In 5.5-day-old embryos, the density of lipid droplets is further reduced. All three major tissues of the embryo are visible at this stage: the distal epiblast, the extraembryonic ectoderm, and covering them both, the visceral endoderm. The proaminotic cavity is also visible (yellow asterisk). The plasma membrane signal is much stronger but not uniformly so, being particularly distinct in the basolateral aspect of cells of the visceral endoderm [blue arrow in Fig. 4c]. The interface between the epiblast end extraembryonic ectoderm is also distinctly visible in THG images. The epiblast has a marked reduction of THG signal (the tissue between the blue arrow and the yellow asterisk). THG signals degrade significantly as we focus deeper in 4.5- and 5.5-day-old embryos (Video 2 shows 3-D opacity rendering of 5.5-day-old embryo). This is largely due to the specimen-induced aberrations,¹⁶ which are not compensated in the images shown here.

3.2.

Long-Term HG Imaging

The combination of adaptive HG microscopy with new culture techniques allowed observation of dynamic changes in the embryo during its development. Video 3 shows 2-h long time-lapse imaging of a single optical section ∼30-μm deep in the blastocyst with 0.2-μm/pixel resolution and at a frame rate of 20 frames/min. One can see lipid droplets of various sizes at high contrast in THG. The trophectoderm and inner cell mass can be clearly seen, and outlines of individual cells within them are faintly visible. One can also very clearly see the blastocoel cavity. The time-lapse sequence shows the blastocoel cavity growing in size, in part, by the incorporation of a smaller cavity near the beginning of the sequence. One can also see lipid droplets moving both within as well as in and out of the image plane.

Video 3

Two-hour time-lapse of a single optical section ∼30 μm deep in the blastocyst. The frame rate of this movie is set to 1 frame/min. (QuickTime, 4.3 MB)

We also performed 3-D time-lapse HG imaging of the 2.5-day-old embryos (precompacted morulae) overnight. Figure 5 shows optical sections from a 19-h-long imaging procedure (with lateral resolution 0.5 μm/pixel, 11 optical sections 3 μm apart with a 1-h interval between each 3-D image stack). Several embryos were imaged simultaneously from morula to blastocyst stage. The wide field of view of the microscope was essential in this case so that we were able to image 9–12 embryos at a time. The embryos develop normally during the course of imaging, which ranged 18–40 h (i.e., able to compact, form cyst, and hatch from zona pellucida). Video 4 shows a 3-D movie, and and Video 5 is an optical section over 19 h, with a clearly formed cyst marked with a yellow asterisk. The cell membrane was obscured by small and numerous lipid droplets while they are more pronounced in older blastocysts, whose number of small lipid droplet had reduced (Fig. 5, 0 h and 18 h, respectively).

Fig. 5

Optical sections from time-lapse imaging of mouse embryos showing morula compaction (x and x^′) and blastocyst (asterisk) formation. Scale bar is 50 μm. See Video 4 for animated 3-D time-lapse movie and Video 5 showing a time-lapse sequence of a single optical section. (Color online only.)

Video 4

3-D time-lapse of the 2.5-day-old embryos over 19 h (lateral resolution 0.5 μm/pixel, 11 optical sections 3 μm apart with 1-h interval between each 3-D image stack). X, X ^′, X ^″ marks an embryo at different stages of development; precompacted morula, morula, and blastocyst, respectively. (QuickTime, 202 KB)

Video 5

Time lapse of embryo development (a single optical section) from precompacted morula to blastcyst during 19-h long THG imaging (QuickTime, 767 KB)