Probing Ordered Lipid Assemblies with Polarized Third-Harmonic-Generation Microscopy

Ordered lipid assemblies are responsible for important physiological functions including skin barrier and axon conductivity. However, techniques commonly used to probe molecular order such as X-ray scattering and nuclear magnetic resonance are not suited for in-situ tissue studies. Here, we identify and characterize a novel contrast mechanism in nonlinear optical microscopy which is sensitive to molecular ordering in multilamellar lipid vesicles (MLVs) and in samples obtained from human skin biopsy: polarized third-harmonic generation (P-THG). We develop a multiscale theoretical framework to calculate the anisotropic, nonlinear optical response of lipid arrays as a function of molecular order. This analysis reveals that conserved carbon-carbon bond and aliphatic tail directionality are the atomic- and molecular-scale sources of the observed P-THG response, respectively. Agreement between calculations and experiments on lipid droplets and MLVs validates the use of P-THG as a probe of lipid ordering. Finally, we show that P-THG can be used to map molecular ordering in the multilamellar, intercorneocyte lipid matrix of the stratum corneum of human skin. These results provide the foundation for the use of P-THG in probing molecular order and highlight a novel biomedical application of multiphoton microscopy in an optically accessible tissue relevant to monitoring lipid-related disorder.

Popular Summary

Bilayer structures formed by ordered assemblies of the chemically “bigamous” lipid molecules (which have chemical affinity for both water and oil) are building blocks found in all mammalian biological cells. In particular, the proper formation of stacked lipid bilayers plays a key role in the function of skin as a protective barrier, and in the electrical connectivity of neurons. Probing and measuring the lipid-molecular ordering in biological tissues, or in situ, and correlating that knowledge with the functions of the tissues is therefore scientifically, and even medically, important. In this study, we establish a novel method for characterizing in-situ lipid ordering in human skin samples with subcellular 3D resolution. We do so by exploiting and refining a nonlinear (multiphoton) microscopy technique that depends on third-harmonic generation and on light polarization.

Ordered lipid structures are in fact nonlinear optical materials. When illuminated by an optical (laser) beam, they interact with the incident photons and emit photons with multiples of the energy of the original photons. Third-harmonic generation corresponds to the photon emission of tripled energy or frequency and is the basis of our method. As we have discovered through a theoretical analysis that spans from the molecular scale to the lipid-assembly scale, the light resulting from third-harmonic generation in a lipid-bilayer assembly is actually sensitive to how the molecules (or more specifically the chemical bonds in the molecules) are oriented with respect to the polarization of the excitation beam. Since ordering of lipid molecules in a supramolecular assembly is intimately related to their orientations in the assembly, the observable sensitivity to molecular orientation translates into an observable sensitivity to the molecular ordering. Based on this theoretical understanding, we are able to experimentally image, with micrometer resolution, lipid orientation and order not only in artificial lipid assemblies, but also in tissue samples obtained from human skin biopsy. Compared to the existing techniques for characterizing lipid order, such as nuclear magnetic resonance, the optical method represents a huge improvement in spatial resolution, can be used in situ, and has the advantage of not requiring the use of unnatural bulky probes.

Our proof-of-principle study demonstrates the promise of polarization-sensitive third-harmonic-generation microscopy for probing molecular order in biological tissues. We hope that further development of the method, by us and by others, will turn it into a practical method for physiological studies of ordered molecular assemblies.

Figure 1

Visual summary of calculations. (a) In the bond-additivity model, each bond contributes ${\gamma }^{\mathrm{mol}}$. The non-CC bonds contribute isotropically while the CC-bond directions give rise to a directed ${\chi }^{(3)}$. The nonlinear dipole moment $|e{\mathbf{r}}^{(3)}(3\omega )|$ is calculated as a function of excitation angle [Eq. (1)] for ${\gamma }^p$ (blue mesh), ${\gamma }^i$ (green mesh), and ${\gamma }^{\mathrm{DMPC}}$ (red mesh). The black lines are the same length and indicate the molecular z axis. (b) The lipids within the bilayer rotate freely about $\psi$, $\varphi$ and adopt a Gaussian distribution of angles relative to the bilayer normal (laboratory X axis). Averaging of ${\gamma }^{\mathrm{DMPC}}$ about these axes results in ${\chi }^{(3)\mathrm{ML}}$. The grey mesh shows the lipid nonlinear polarization $|{\mathbf{P}}^{(3)}(3\omega )|$, and the black line indicates the bilayer normal. (c) Left: Simulated focal volume half-filled with ordered lipid (dark gray) and either vacuum or water (clear). Right: Focal distribution of induced $\mathbf{P}(3\omega )$ in each material at the focus ($\mathrm{NA}=0.8$, $\lambda =1180\mathrm{nm}$, $n_{\omega }=n_{3\omega }=1.33$, scale bar 500 nm).

Figure 2

Calculated nonlinear optical properties of multilamellar lipid structures and interfaces. The ${\chi }^{(3)\mathrm{ML}}$ tensor elements of (a) liquid-phase $(⟨\theta ⟩=0°)$ and (b) gel-phase $(⟨\theta ⟩=30°)$ lipids as a function of disorder $\sigma$. Higher ordering results in greater ${\chi }^{(3)\mathrm{ML}}$ anisotropy for the liquid phase; in the gel phase, optical isotropy is achieved at lower disorder. The elements are also shown to approach the same value of $⟨{\chi }^{(3)\mathrm{ML}}⟩$ (dotted line) as the system approaches isotropy. (c),(d) The anisotropy coefficients ${\delta }_X$, ${\delta }_Y$ characterize the magnitude of the anisotropy-induced P-THG signal oscillation. Generally, for a pair of coefficients, ${\delta }_X>0$ and ${\delta }_Y<0$, and their magnitude is larger for liquid-phase interfaces (gray curves) relative to the gel-phase interfaces (purple curves). For an interface between two isotropic media, ${\delta }_X={\delta }_Y=0$; thus, the gel-phase curves [corresponding to the tensor elements in (b)] cross at $\sigma \approx 45°$, and the liquid-phase curves [corresponding to (a)] cross at $\sigma \approx 75°$. Comparing the (c) vacuum-lipid and (d) water-lipid interfaces shows that ${\chi }^{(3)}$ matching of the second, isotropic material can potentially be used to enhance the P-THG contrast.

Figure 3

Calculated P-THG response as a function of lipid order parameters. (a) The components of ${\chi }^{(3)\mathrm{ML}}$ are determined by the physical distribution of lipid angles relative to the X axis. A low angle and low $\sigma$ correlate with the highest P-THG modulation. The arrow points from the highest-modulation parameters to lowest. (b) Calculated modulation amplitude as a function of lipid parameters and interface type. Modulation is generally higher at lipid-water (blue curves) than at vacuum-lipid (grey curves) interfaces, showing the increased sensitivity to anisotropy due to the partial ${\chi }^{(3)}$ matching. The modulation is also consistently larger for the liquid-phase lipids (closed symbols) relative to the gel-phase (open symbols) lipids for a particular interface type. Experimental values for the modulation in the MLV and lipid droplet (LD) are indicated with dashed lines. (c),(d) The calculated angular profiles for four multilamellar lipids, color-coded to match (a). The P-THG signal is lowest when the excitation field is parallel to the bilayers. The fitting curves are of the form $I=A+B\cos (\Omega {)}^2$.

Figure 4

Setup used for the various imaging experiments. $\mathrm{OBJ}=\mathrm{\text{Objective}}$, $\mathrm{PMT}=\mathrm{\text{photomultiplier}}$ tube. The excitation path is represented by the red beam; blue beams indicate both the forward and the epi (backward) emission pathways. THG and CARS images are usually detected in the forward direction, while the 2PEF images are epidetected using a dichroic mirror. The general form of this setup is typical for any nonlinear optical microscope, with the inclusion of a polarizer and a half wave plate to guarantee accurate control of the polarization state at the focus.

Figure 5

P-NLOM imaging of ordered and amorphous lipid structures. (a) THG and CARS images of MLVs generated with two orthogonal polarizations (scale bar $15\mu \mathrm{m}$). The maximum CARS signal is obtained with excitation fields polarized parallel to the ${\mathrm{CH}}_2$ moieties, i.e., to the MLV surface. Two types of THG signals are detected. The interface THG signal (analyzed in this study) is maximized for an excitation field polarized parallel to the CC bonds (i.e., to the surface normal) and retains the excitation polarization. The bulk THG signal is due to MLV birefringence and is polarized perpendicular to the excitation. (See also movie S1 in Ref. 32.) (b) Comparison between the polarization dependence of THG from the surface of a lipid droplet (isotropic lipids) and a MLV (with approximately ${10}^3$ bilayers), clearly showing the sensitivity of P-THG to molecular ordering. The MLV data analyses indicate that pixel averaging is not necessary to retrieve accurate estimates of the interface angle and modulation amplitude. (c)–(e) Extraction of order parameters. On a per-pixel basis, the fits can be used to construct maps of (c) the lipids orientation and (d) modulation amplitude based on the interface THG signal (scale bar $15\mu \mathrm{m}$). The same fitting procedure can accurately extract physical parameters in more complex systems, such as the large lipid aggregate in (e) (scale bar $15\mu \mathrm{m}$; see also movie S2 in Ref. 32).

Figure 6

Nonlinear optical imaging of samples from human skin biopsies. (a) Within the skin stratum corneum (labeled “SC”), stacked bilayers exist between the dead corneocytes (labeled “c”). Vertically oriented folds extending tens of microns deep into the epidermis (labeled “e”) provide an optimal geometry for P-THG imaging (dashed blue line). (b) THG images of the skin highlight cell nuclei (labeled “n”) as well as the folds (“f”) (scale bar $50\mu \mathrm{m}$). (c)–(e) Subregion of the rectangular area indicated in (b). In human skin biopsies, endogenous 2PEF from the corneocytes alternating with THG signals (see white arrowheads) localizes the THG response to the intercellular lipid interfaces. (f),(g) THG images of a human skin biopsy approximately $25\mu \mathrm{m}$ from the skin surface generated using two different polarizations, indicated by the straight doubleheaded arrows (scale bar $50\mu \mathrm{m}$). Note how the folds that are oriented perpendicular to the excitation polarization appear brighter than the ones with parallel orientations. As with the MLV, the P-THG signal can be fitted to extract the stacked lipids orientation (h) and modulation (i), to gain insight into the lipid ordering within the interfaces. Even within the tissue, a strong modulation is observed exclusively from ordered phases and not from other structures. See also the movie S3 in Ref. 32.

Figure 7

Calculation of the signal modulation using the fast, approximate method. (a) Modulation as a function of the disorder parameter, similar to the predicted trend from the full numerical simulations. (b) Relative error between the two predictions, showing that the fast calculation is consistently within 10% (although generally even lower) of the numerical simulations.