Raman microscopy is an emerging molecular imaging technique that does not require the labeling of targeted molecules in specimens. A key to the emergence of Raman microscopy is the development of high-speed imaging techniques that enable the analysis of biological specimens susceptible to temporal changes or alterations. Here, we introduce a high-speed Raman hyperspectral imaging technique developed based on line-illumination Raman microscopy. The technique, namely multiline-illumination Raman microscopy, equips periodic arrays of line-shaped laser focus and confocal slits, enabling simultaneous spatial sampling of 10,000-20,000 Raman spectra and spectral sampling of 100 wavenumbers per exposure, which performs high-throughput Raman hyperspectral image acquisition of (105 spectra × 102 wavenumbers)/min. We anticipate that our technique will enhance the utility of Raman microscopy in practical applications in biology and medicine.
Introduction
Raman microscopy has emerged in the last decade as a molecular imaging technique enabling label-free molecular analysis [1]. The capability of Raman microscopy has been studied for a variety of biomedical applications, i.e., observing drug response in living cells at the molecular level [2], cell sorting [3], intraoperative rapid diagnosis [4], and identification of cell differentiation [5]. However, the slow imaging speed of Raman microscopy hampers the examination of practical applications of Raman microscopy in the life sciences, mainly due to the small Raman scattering cross-section and point-by-point scanning scheme for image construction. Although multiplexing spatial sampling, e.g., by line-illumination [6] and multifocal illumination [7], has been applied to increase the speed by 2-3 orders of magnitude, Raman hyperspectral imaging often takes a few tens of minutes or even longer than an hour [6, 8].
Strategy for accelerating Raman imaging
In conventional confocal Raman microscopy, a laser beam in the form of a spot illuminates the sample, and a single Raman spectrum is recorded using sensors only in several lines of a two-dimensional sensor array (Fig. 1A). In multiplex spatial sampling technique using line-shaped illumination, the spatial distribution of the spectra in the sample is recorded on the sensor array by expanding the spatial information along y-axis, which is orthogonal to the wavelength dispersion axis (x-axis), as shown in Fig. 1B. The number of spectra simultaneously acquired here is determined by the number of sensors in the y-axis used for recording, i.e., typically 100-1,000, resulting in the imaging speed faster by two or three orders of magnitude compared to confocal Raman microscopy [9]. To further improve the imaging speed by multiplex spatial sampling, we increased the number of illumination lines in the x-axis to acquire more spatial information simultaneously in a trade-off with reducing the wavenumber information (Fig. 1C).
Multiline-illumination Raman microscopy with slit array detection
We have developed a multiline-illumination Raman microscope equipped with a confocal slit array (Fig. 2) [10]. A cylindrical lens array formed a high-power laser beam with a wavelength of 532 nm into multiple line-shaped laser beams. The multiple line-shaped beams were relayed to the sample plane and excited Raman scattering in a sample. The Raman scattering was collected with an objective lens, separated from the excitation beam path by using an edge filter, and refocused at the entrance of the spectrophotometer, where a periodic slit array was located instead of an ordinary entrance single slit. Each slit was conjugated to one of the irradiated regions at the sample and worked as a confocal slit. To avoid unwanted spectral mixtures among different sample regions in the sensors of the spectrophotometer, the wavelength range of Raman scattering entering the spectrophotometer was limited by a bandpass filter. A comb-like Raman hyperspectral image excited by multiple illumination lines is acquired by a single sensor exposure, where each Raman spectrum includes specific Raman peaks from target molecules in the sample.
Raman hyperspectral image from polystyrene (PS) and polymethylmethacrylate (PMMA) beads dispersed on a quartz coverslip is shown in Fig. 2. In the spectra corresponding to the CH vibration region (2,850–3,100 cm-1), the characteristic Raman peaks of PMMA (green) and PS (magenta) are found at 2,953 and 3,047 cm-1, respectively. To acquire a two-dimensional Raman hyperspectral image (Raman spectra distributed in two-dimensional space), the sample was scanned with the slit array and multiple illumination lines by steering a galvanometer mirror. A shutter in the beam path opened and closed synchronously with the start and end of a sensor frame exposure, respectively, to prevent the excitation laser beams from excessively irradiating the sample.
High-throughput Raman hyperspectral imaging of cell and tissue
Here, we present Raman images of biological specimens to show the utility of the multiline-illumination Raman microscope for molecular imaging of biological samples. Figure 3A displays the Raman spectra and image for living cultured HeLa cells. In the spectra, the band at 2,854 cm-1 can be assigned to the CH2 asymmetric stretching vibrational mode, while the band at 2,935 cm-1 can be assigned to the CH3 symmetric stretching vibrational mode. The Raman images reconstructed using the intensities at 2,854 and 2,935 cm-1 can illustrate the distributions of lipid and protein, respectively, in the cells [11]. Small dots of lipid, i.e., lipid droplets, are imaged without significant distortion, indicating that the multiline-illumination Raman microscope has sufficiently high spatial resolution to observe organelles in living cells.
A measurement result for clinical specimens of frozen dermal tissue is shown in Fig. 3B. In the spectra, two characteristic Raman bands were identified at 2,850 and 2,945 cm-1. The Raman image reconstructed with the intensity at 2,945 cm-1 shows collagenous fibrous structures that occupy most of the part in the dermal tissue [12], while the Raman image reconstructed at 2,850 cm-1, exhibiting a warped shape, presumably shows elastin fibers that connect collagen structures which distribute among them [13]. These results indicate that the multiline-illumination Raman microscope visualizes tissue structures that can be useful for studies of tissue chemistry and morphology in a label-free manner [14,15].
Each Raman hyperspectral image in Fig. 3 is composed of 920 × 1,300 spectra. The data acquisition time was 8.0 min (5.0-s exposure and 0.2-s readout per single frame acquisition of the sensors) for each dataset, corresponding to a throughput of 2 × 105 spectra/min. This throughput is an order of magnitude higher than that in line-illumination Raman microscopy with single-slit detection, which can realize the ultrafast Raman imaging of biological specimens in practical conditions.
Conclusion
We have developed a high-speed line-illumination Raman microscope equipped with a confocal slit array, where the image acquisition speed was improved by increasing the number of spectra that can be acquired with a single exposure. With the developed technique, we achieved a throughput with the order of 105 spectra/min in cell and tissue imaging, marking a 10-fold improvement in speed compared to the single-line-illumination technique. Currently, to accelerate the image acquisition further, we are developing a technique capable of achieving speeds on the order of 106 spectra/min by efficiently detecting Raman scattering in a shorter exposure time. We believe that the presented high-speed molecular imaging technique will further expand the utility of Raman microscopy in the fields of biology and medicine, not only for pursuing basic science but also for developing new applications in clinics and related industries.
Contact
Prof. Dr. Katsumasa Fujita
Department of Applied Physics
Graduate School of Engineering
Osaka University, Japan
ORCID: 0000-0002-2284-375X
[email protected]
References
[1] Krafft, C. et al. (2017). Label-Free Molecular Imaging of Biological Cells and Tissues by Linear and Nonlinear Raman Spectroscopic Approaches. Angewandte Chemie International Edition. DOI: 10.1002/anie.201607604.
[2] Li, M. et al. (2022). Label-free chemical imaging of cytochrome P450 activity by Raman microscopy. Communications Biology. DOI: 10.1038/s42003-022-03713-1.
[3] Nitta, N. et al. (2020). Raman image-activated cell sorting. Nature Communications. DOI: 10.1038/s41467-020-17285-3.
[4] Hollon, T.C. et al. (2020). Near real-time intraoperative brain tumor diagnosis using stimulated Raman histology and deep neural networks. Nature Medicine. DOI: 10.1038/s41591-019-0715-9.
[5] Schulze, H.G. et al. (2010). Assessing Differentiation Status of Human Embryonic Stem Cells Noninvasively Using Raman Microspectroscopy. Analytical Chemistry. DOI: 10.1021/ac902697q.
[6] Palonpon, A.F. et al. (2013). Raman and SERS microscopy for molecular imaging of live cells. Nature Protocols. DOI: 10.1038/nprot.2013.030.
[7] Okuno, M. and Hamaguchi, H. (2010). Multifocus confocal Raman microspectroscopy for fast multimode vibrational imaging of living cells. Optics Letters. DOI: 10.1364/OL.35.004096.
[8] Kumamoto, Y. et al. (2019). High-Throughput Cell Imaging and Classification by Narrowband and Low-Spectral-Resolution Raman Microscopy. The Journal of Physical Chemistry B. DOI: 10.1021/acs.jpcb.8b11295.
[9] Kumamoto, Y. et al. (2022). Slit-scanning Raman microscopy: Instrumentation and applications for molecular imaging of cell and tissue. Journal of Applied Physics. DOI: 10.1063/5.0102079.
[10] Mochizuki, K. et al. (2023). High-throughput line-illumination Raman microscopy with multislit detection. Biomedical Optics Express. DOI: 10.1364/BOE.480611.
[11] Xu, Y. et al. (2022). Phenotype Identification of HeLa Cells Knockout CDK6 Gene Based on Label-Free Raman Imaging. Analytical Chemistry. DOI: 10.1021/acs.analchem.2c00188.
[12] Ogura, Y. et al. (2019). Texture analysis of second-harmonic-generation images for quantitative analysis of reticular dermal collagen fibre in vivo in human facial cheek skin. Experimental Dermatology. DOI: 10.1111/exd.13560.
[13] Zhuo, S. et al. (2006). Multimode nonlinear optical imaging of the dermis in ex vivo human skin based on the combination of multichannel mode and Lambda mode. Optics Express. DOI: 10.1364/OE.14.007810.
[14] Mostaço-Guidolin, L.B. et al. (2013). Collagen morphology and texture analysis: from statistics to classification. Scientific Reports. DOI: 10.1038/srep02190.
[15] Becker, L. et al. (2023). Raman microspectroscopy identifies fibrotic tissues in collagen-related disorders via deconvoluted collagen type I spectra. Acta Biomaterialia. DOI: 10.1016/j.actbio.2023.03.016.
