2.1.

Cannabis sativa L. and ${\mathrm{\Delta }}^9$-Tetrahydrocannabinolic Acid

Samples of dried female Cannabis sativa L. plants, consisting primarily of pistillate flowers and adjacent bracts, were obtained from a local distributor in Enschede, The Netherlands. White-light reflection stereomicroscopy was used to verify the presence and density of glandular trichomes and to identify structures of interest. Small sections of the pistils were removed from the central flower, with great care taken to ensure that the regions to be imaged were not damaged. Small leaves were likewise separated from the flower at the bases of their stems. These samples were immersed in deionized water to mitigate thermal effects and sealed between two clean #1 cover glasses. Slight morphological changes of the samples due to rehydration were observed, but did not affect the CARS measurements. The distributions and concentrations of cannabinoids within the samples were not expected to be modified because of the hydrophobicity of the cannabinoids.

For reference spectra we used $>97\%$ purity (as measured with HPLC and ${}^1\mathrm{H}\text{-}\mathrm{NMR}$) crystalline THCa (39382-10MG; Sigma Aldrich), which was stored at $-20°\mathrm{C}$. Immediately before imaging we sealed a small amount of THCa in air between two clean #1 cover glasses. Care was taken to avoid contamination. New samples were prepared for each imaging session.

2.2.

Optical Setup

Our optical setup has been described in depth elsewhere.¹³ Briefly, a frequency-doubled $\mathrm{Nd}:{\mathrm{YVO}}_4$ laser (Paladin; Coherent, Inc.) synchronously pumped an OPO (Levante Emerald; APE Berlin GmbH), producing two dependently tunable near-IR beams. For imaging the Cannabis plants we used the signal beam from the OPO, nominally centered around 810 nm, combined with the laser fundamental at 1064 nm to excite high-frequency alkyl vibrations near $3000{\mathrm{cm}}^{-1}$. To obtain spectra of the pure THCa crystals we used the OPO signal and OPO idler beams together, nominally centered at 920 and 1260 nm, respectively. The latter pair provided up to $350{\mathrm{cm}}^{-1}$ of tuning bandwidth, twice that of the signal-1064 combination. The selected pair of beams was sent collinearly into an inverted laser-scanning microscope (IX71/FV300; Olympus) and focused into the sample with a $60\times 1.2\mathrm{NA}$ water immersion objective (UPLSAPO 60XW/IR; Olympus). Transmitted CARS signals were collected with a long-working-distance 0.55 NA objective (MWUCD; Olympus), spectrally filtered to reject the pump and Stokes beams (Chroma HQ660/40 for the plants, Semrock FF01-716/43 for the pure THCa), and detected on a photomultiplier tube (R3986; Hamamatsu). Back-scattered CARS signals were reflected off of a dichroic mirror (FF775; Semrock), and filtered and detected with identical but separate components as the transmitted CARS signal.

Two-photon-excited autofluorescence from the wide variety of native fluorophores in the plants can be a problem in CARS experiments. Chlorophyll a, for example, has a very strong absorption band in the blue, but only weakly absorbs in the green. For plant imaging we used the OPO signal ($\sim 820\mathrm{nm}$) as the pump beam in the CARS process and the laser fundamental (1064 nm) as the Stokes beam. The second harmonic of the pump beam, around 410 nm, is strongly absorbed by chlorophyll a, while the sum-frequency of the pump and Stokes beams at 470 nm and the second harmonic of the Stokes beam at 532 nm are not. To separate these processes we switched the laser fundamental beam on and off at high frequency (2.4 MHz) with an acousto-optic modulator. The light emitted from the sample then contained two components: a fluorescent DC component from two-photon absorption of the pump beam and an AC component at 2.4 MHz with the CARS signal and fluorescence from the other two-photon absorptions, both of which were recorded on the PMT. The output of the detector was sent to a high-speed lock-in amplifier (HF2LI; Zurich Instruments), where the large DC component was filtered out, leaving only the CARS signal and residual fluorescence. We used a high modulation frequency and short time constant (5 μs) on the lock-in amplifier to maintain a fairly high imaging rate.

For hyperspectral CARS imaging we scanned the laser beams over the sample to create a two-dimensional image (frame) at one vibrational frequency, then slightly changed the orientation of a Lyot filter inside the OPO cavity between frames to shift to the next frequency. The minimum frequency shift is $0.7{\mathrm{cm}}^{-1}$, though for speed considerations we used $2{\mathrm{cm}}^{-1}$ intervals. Many iterations of this process generated the full hyperspectral datacube. An image processing algorithm projected this three-dimensional datacube into a color-coded two-dimensional image wherein each spectrum is uniquely and qualitatively represented. Specific regions of interest were then selected based on visual inspection of these projections, from which CARS spectra were extracted from the original datacube. These spectra were corrected in post-processing for variations in pump and Stokes beam power.

3.1.

${\mathrm{\Delta }}^9$-Tetrahydrocannabinolic Acid

To determine the spectral markers for locating THCa in our experiments in Cannabis sativa L. plants, we first obtained reference spectra of pure crystalline THCa. A representative hyperspectral image projection is shown in Fig. 1(a), with the spectra of the indicated regions shown in Fig. 1(b). This image of $256\times 256$ pixels at 117 spectral points was recorded in 2 min. Surprisingly, we find two unique spectra associated with pure THCa, which we call ${\mathrm{THC}}_1$ and ${\mathrm{THC}}_2$. The ${\mathrm{THC}}_1$ spectrum (blue region and curve) contains many overlapping alkyl vibrational modes, with four major peaks located at 2825, 2868, 2907, and $2940{\mathrm{cm}}^{-1}$, while the ${\mathrm{THC}}_2$ spectrum (red region and curve) has a single prominent feature at $2903{\mathrm{cm}}^{-1}$ and a minor secondary peak at $2960{\mathrm{cm}}^{-1}$. Based on the structure of the THCa molecule we expect a complicated vibrational spectrum with significant contributions from both ${\mathrm{CH}}_2$ and ${\mathrm{CH}}_3$ modes, which we can reconcile with ${\mathrm{THC}}_1$ (blue curve), although specific band assignments are not attempted. The ${\mathrm{THC}}_2$ spectrum (red curve) is especially interesting because this spectrum was represented in approximately half of the THCa crystals that we measured. Both spectra are insensitive to the orientations of the crystals relative to the polarizations of the laser beams, indicating that these two spectra do not arise from a systematic measurement artifact, and the crystals of ${\mathrm{THC}}_1$ and ${\mathrm{THC}}_2$ are indistinguishable under white-light illumination. From these observations and the HPLC and ${}^1\mathrm{H}\text{-}\mathrm{NMR}$ data, we propose that these two spectra correspond to two distinct crystal solid-state forms of THCa. Large variations in the vibrational spectra between multiple crystal forms of the same compound are not unheard-of, as has been shown with spontaneous Raman for mannitol²¹ and lactose.²²

Fig. 1

(a) Hyperspectral CARS image of pure THCa crystals displaying two distinct colors. (b) Spectra extracted from the indicated regions of (a). Key marker frequencies for ${\mathrm{THC}}_1$ are 2825, 2860, 2905, and $2940{\mathrm{cm}}^{-1}$, while ${\mathrm{THC}}_2$ has only a single strong peak at $2900{\mathrm{cm}}^{-1}$. Small inclusions of ${\mathrm{THC}}_2$ within the blue region generate additional intensity at $2900{\mathrm{cm}}^{-1}$ in the ${\mathrm{THC}}_1$ spectrum. The THCa structure is shown in part (b). Note that both spectra in part (b) have been normalized to the maximum value of the ${\mathrm{THC}}_2$ spectrum, and that the ${\mathrm{THC}}_1$ spectrum has been scaled by an additional factor of three.

3.2.

Cannabis sativa (Leaf)

To measure THCa within the leaves of a Cannabis plant we detect the back-scattered CARS signal, since strong absorption of the 650-nm anti-Stokes emission by chlorophyll and high losses due to scattering in the leaves precluded detection in transmission. The presence of numerous fluorescent bodies necessitated use of modulated detection. A representative hyperspectral image, consisting of $512\times 512$ pixels over 80 spectral points and imaged at 7 s per frame, is shown in Fig. 2(a), with spectra of the indicated regions in Fig. 2(b).

Fig. 2

Hyperspectral CARS image (a) and spectra (b) from a Cannabis sativa L. leaf, recorded in reflection, showing native fluorescence (green region and curve) and a ${\mathrm{THC}}_1$ CARS signal (blue region and curve).

Immediately obvious from Fig. 2 is that the two highlighted regions have very dissimilar spectral profiles. The region indicated in blue is spectrally very close to that of pure ${\mathrm{THC}}_1$, while the region outlined in green shows the flat, featureless spectral curve typical of fluorescence. We believe that this fluorescence is mostly from chlorophyll b, which can be excited by the simultaneous absorption of one pump photon and one (modulated) Stokes photon and whose emission is therefore modulated.

The physical morphology also differs between the ${\mathrm{THC}}_1$ and fluorescent regions, with the former appearing spatially more homogeneous and having smoother edges than the latter. The spatial features of the ${\mathrm{THC}}_1$ signal appear generally consistent with a storage vesicle or reservoir. Small irregularities of the edges of this region may indicate a rupture of the vesicle that has allowed the ${\mathrm{THC}}_1$ to diffuse into the surrounding tissue and deformed the walls of the vesicle. The lack of other spectral signatures in the ${\mathrm{THC}}_1$ region points toward a relatively high purity and local concentration of cannabinoids, and the absence of a discernible microscopic structure indicates either a solvated or nano-crystalline phase.

3.3.

Cannabis sativa (Pistil)

The pistils of flowering Cannabis plants are known to contain large quantities of cannabinoids, with average concentrations of 10 to 12%/w (see Ref. 17). We have observed that, unlike the leaves, pistils contain very dense distributions of cannabinoids in complicated heterogeneous structures. Because the pistils are fairly thin and their glandular trichomes are nearly transparent, CARS detection in transmission is utilized. This confers two advantages: CARS emissions are highly directional, with a strong preference for scattering forward in a relatively narrow cone, and so are efficiently collected and relayed to detectors in that direction; and CARS emissions coherently add as the beams propagate through a sample, by which process the signals from large objects (those larger than the wavelength of light) benefit tremendously.

Most of the THCa signals that we have recorded appear to originate in structures with linear dimensions greater than 1 μm, as illustrated by the $z$-stack series in Fig. 3. To create this image the laser focus was axially translated by 1 μm through the sample between individual frames. A total of 60 frames were recorded at a constant vibrational frequency of $2940{\mathrm{cm}}^{-1}$. Each frame was recorded in 10 s, including three averages. We verified that the bright signals correspond to a ${\mathrm{THC}}_1$ CARS signals by taking a hyperspectral image (not shown) at a slice in the center of the stack. To depict the three-dimensional structure of the entire trichome we have applied a $z$-dependent color to each slice (from the $Z$-Stack functions of the MBF library in ImageJ). The hazy background in the top half of the image is the nonresonant background of the water immersion medium. Reds and yellows indicate regions toward the bottom of the stack, while blues and purples are at the top of the stack. There are four distinct regions that can be identified. The red/orange section at top right is highly structured and dense, and contains small voids or vacuoles. Directly beneath that section are two unconnected regions of similar morphology at different heights in the sample, colored yellow/green and light blue, respectively. These regions contain higher concentrations of cannabinoids, but in elongated and narrow structures. The left side of the trichome displays a large and more diffuse distribution without clear and sharp boundaries (yellow, green, light blue), similar to that observed in the bract. Finally, the far top right of the trichome has a number of very thin regions of cannabinoids, shown in green, that may be closely associated with or bound to cell membranes. From these images the compartmentalized distribution of the cannabinoids is readily apparent. The observation of these different morphologies has not, to the best of our knowledge, been previously reported.

Fig. 3

Color-coded projection of a CARS $z$-stack in a glandular trichome, measured at the ${\mathrm{THC}}_1$ peak at $2940{\mathrm{cm}}^{-1}$. Reds and oranges indicate the bottom of the stack, while blues and purples are at the top of the stack. Many different physical morphologies are visible at different regions within this trichome.

We have also imaged a region within an intact stigma of a female flower that contains both ${\mathrm{THC}}_1$ and ${\mathrm{THC}}_2$ in close proximity. The hyperspectral CARS image shown in Fig. 4 was recorded at a depth of about 30 μm, with $512\times 512\times 80$ pixels at 10 s per frame. The spectra from the magenta and blue regions of the image correspond approximately to ${\mathrm{THC}}_1$ and ${\mathrm{THC}}_2$, respectively, and are shown in Fig. 4(b). Additionally, a small volume of a native lipid, which is easily identified by its characteristic ${\mathrm{CH}}_2$ symmetric stretch at $2845{\mathrm{cm}}^{-1}$, is outlined in green. Deviations from the reference spectrum of ${\mathrm{THC}}_1$ can be attributed to the presence of other molecules in the focal volume of the laser beam, including other cannabinoids such as cannabinol and cannabidiol.¹⁴ The broadening of the ${\mathrm{THC}}_2$ peak is a result of detector saturation, a side effect of increasing the contrast of other regions in the image. At nonsaturated gain settings the spectrum is nearly identical to that shown in Fig. 1(b) for ${\mathrm{THC}}_2$.

Fig. 4

Hyperspectral CARS image (a) and CARS spectra (b) of the interior of a stigma showing both representative THCa spectra (magenta and blue regions and curves) as well as native lipids (green region and curve). The presence of many compounds other than THCa in this section of the plant yields a complex mixture, from which it is impossible to find a pure THCa spectrum in any region of interest.