Lipid droplets (LD) are increasingly seen as complex organelles in their own right, and not merely as inert bodies meant simply as energy stores. They are composed of a core of neutral lipids enveloped by a phospholipid monolayer, but also contain a wide variety of proteins, both within the core and embedded in the phospholipid monolayer (reviewed in [1–4]). Fatty acids in LD are used for the generation of energy, membrane synthesis, production of signalling molecules and modification of proteins. Because of this, LD are often found in association with organelles linked to lipid metabolism such as mitochondria, endoplasmic reticulum (ER), endosomes, and peroxiosomes . LD are also involved in several pathological conditions in humans - excess lipid is associated with atherosclerosis  and is characteristic of cancer cells . Infection by Dengue or hepatitis C virus leads to an increase in LD number due to the LD having been commandeered for viral particle production [8–10].
LD are dynamic organelles and change shape, volume and location constantly. Evidence from Drosophila embryos and mammalian cell-lines indicates that microtubules are required for directional movement of lipid droplets [11–13]. LD have been reported to coalesce into larger droplets in a microtubule dependent manner , though this poses questions regarding the volume/surface area relationship between the constituent droplets and the final LD . Uncertainty remains regarding whether LD actually fuse, as some groups have been unable to observe it occurring .
In mammalian embryos, LD are considered primarily to be an energy source, similar to yolk in non-mammalian eggs. However, proteomic approaches in Drosophila suggest that LD might also act as protein-storage organelles in embryos since they contain abundant levels of histones, globular actin, ribosomal subunits and mitochondrial proteins . A similar diversity of proteins has been detected in LD from rat hepatocytes . LD are a dominant feature of pre-implantation embryos from the pig and cow but are also present in mouse and human embryos [17–19]. Studies in mouse embryos have shown an association of LD with organelles such as autophagosomes and mitochondria , but in general, little is known about the function and behaviour of LD in mouse embryos.
A major limitation to the study of LD is that most LD dyes work best with fixation, which precludes studying dynamic behaviour. GFP fusion proteins that tag LD have been developed [13, 14], but using these in mouse embryos requires the invasive injection of the fusion construct or the production and maintenance of transgenic lines.
Harmonic generation microscopy (HGM) has great potential for the three-dimensional, label-free imaging of developing embryos as demonstrated with zebrafish  and Drosophila . The HGM is a laser scanning microscope that takes advantage of second and third harmonic generation (SHG and THG respectively), that result in the generation of photons of half and one-third the illumination wavelength respectively. SHG requires non-centro symmetric media and is generated mainly by structures like muscle, organised microtubules in the mitotic spindle, and collagen fibres . THG probes interfaces and optical inhomogeneities - local variations in the third order non-linear susceptibility (χ(3)) and/or refractive index. LD are the major source of THG contrast in cells . HGM has other advantages - emitted photon energy is exactly the same as incident energy so there is no energy deposition in the specimen. It typically employs laser wavelengths in the near infra-red region (900-1500 nm), reducing the absorption and scattering of incident photons, and allowing deeper imaging in thick specimen . HGM is therefore expected to be minimally photo-toxic, making it ideal for studying living embryos.
The performance of microscopes is often compromised by aberrations, arising from imperfections in the optical system or due to physical properties of the specimen. Due to the non-linear dependence of the harmonic signal on the focal intensity, HGM is particularly sensitive to the effects of aberrations. Furthermore, as the illumination wavelengths typically used in HGM are outside the specification of most objective lenses, system aberrations can be significant. The problems caused by aberrations can be overcome using adaptive optics, whereby aberrations are corrected using a dynamic element, such as a deformable mirror . As aberration correction leads to more efficient signal generation, the illumination laser power can be reduced. This is particularly important in reducing phototoxic effects in living specimens, extending the period over which they can be observed.
We have developed an adaptive HGM  and new culture techniques for imaging peri-implantation mouse embryos. We find that the THG signal at these stages is generated predominantly by LD. Time lapse HGM shows that LD are very motile, and collide to form larger aggregates that behave as a single unit without actually fusing. This dynamic behaviour is dependent on both microtubules and microfilaments. Finally, we explore the impact of HGM on embryo viability and find that continuous imaging for short time periods does not compromise viability but extended imaging, even with a lower cumulative energy load, does.