Raman microspectroscopy, which combines high molecular specificity of Raman spectroscopy and high spatial resolution of optical microscopy, has emerged as a unique and powerful method for in vivo and in situ living cell studies at the molecular level. Extensive use of this method, however, has been hindered by slow data acquisition speed that is due to small Raman cross-sections of molecules. In this study, I have developed new linear and nonlinear Raman microspectroscopies, multi-focus confocal Raman microspectroscopy and quantitative coherent anti-Stokes Raman scattering (CARS) microspectroscopy for fast vibrational imaging. I have adopted the multi-focus excitation and the image compression techniques to develop a new scheme of spontaneous Raman microspectroscopy, multi-focus confocal Raman microspectroscopy, in order to improve the data acquisition speed without increasing the excitation laser power at the sample. I have combined CARS microspectroscopy with the maximum entropy method (MEM) in order to convert CARS spectra into the Im[X(3)] spectra whose intensities are linear to molecular concentration.

I also paid attention to the absolute quantitative nature of Raman microspectroscopy. The number of Raman scattered photons is strictly proportional to the number of molecules in the focal volume. If the absolute Raman cross-section of a molecule is known, the concentration of this molecule can be directly determined from the observed number of Raman scattered photons. This absolute quantitative character can be a great advantage in molecular imaging, particularly in molecular imaging of living cells. In this study, I have made a challenge to estimate the numbers/concentrations of biomolecules in living cells by determining their absolute Raman cross-sections.

Quantitative CARS microspectroscopy

Figure 1 (a) shows a typical multiplex CARS spectrum obtained from a budding yeast cell. I employ MEM to extract the amplitude and phase of vibrational resonances from the obtained multiplex CARS spectra. The MEM does not require any a priori knowledge of the vibrational bands contained but still is able to retrieve the phase information on the third-order nonlinear susceptibility X(3), whose imaginary part corresponds to ordinary (spontaneous) Raman spectra. Figure 1 (b) shows the reconstructed Im[X(3)] spectrum of Fig. 1 (a) with MEM. Many vibrational bands in the fingerprint region are clearly found with high signal to noise ratio. It has been confirmed that this spectrum is very similar to the corresponding spontaneous Raman spectrum obtained from a yeast cell. CARS microspectroscopy combined with MEM has realized label-free and quantitative molecular mapping of a single living cell in the fingerprint region (CARS molecular fingerprinting).

CARS molecular fingerprinting enables us to obtain label-free and multi-colour images with high speed. Although the fingerprint region is spectrally congested in a living yeast cell, more than 10 vibrationally resonant Im[X(3)] images by MEM have been successfully extracted. This method has been also applied to a time-resolved study of dynamical processes in a single living yeast cell such as rapid cell death by the laser irradiation and the cellular uptake of surfactant molecules.

Multi-focus confocal Raman microspectrometer

The developed apparatus is schematically shown in Fig. 2. It consists of an inverted microscope, a micro-lens array, a pinhole array, a fiber-bundle and a multichannel spectrometer. The micro-lens array is used to split the illuminating laser into beamlets such that the objective lens produces a pattern of 8 x 8 independent foci at the sample. The role of the pinhole array is to improve the axial spatial resolution and to enhance the ratio of the Raman signal to the background by the confocal effect. This multi-confocal configuration is indispensable for obtaining high contrast images of living cells in which weak Raman signals are easily overwhelmed by the strong background arising from the cover glass and the cell culturing medium. The 64 fiber-bundle is arranged in a 8 x 8 rectangular pattern at the collection end and a 1 x 64 linear stack at the detection end, which is connected to the entrance slit of the spectrograph directly. Using this fiber-bundle, a 3-D (2-D spatial and spectral) data cube is converted into a 2-D data array, which can be detected simultaneously by a 2-D CCD camera. The lateral and axial spatial resolutions of the system are determined to be 300 nm and 1 μm, respectively. In the present setup, 48 spectra from 48 points (X: 8 points Y: 6 points) are detectable with one CCD exposure.

Multi-vibratioal mode Raman images obtained at 1446 cm(-1) (Fig. 3 (a)), 1583 cm(-1) (Fig. 3 (b)), 1602 cm(-1) (Fig. 3 (c)) and 1655 cm(-1) (Fig. 3 (d)) are shown in Figs. 3. The Raman bands at 1446 cm(-1), 1583 cm(-1) and 1655 cm(-1) are assigned to the CH bend, the porphyrin in-plane C=C stretch mode of the porphyrin skeleton and the superposition of the cis-C=C stretch of unsaturated lipid chains and the amide I mode of proteins, respectively. The Raman band at 1602 cm(-1) is called the "Raman spectroscopic signature of life". These images show similarity and variation, originating from distributions of different chemical species giving rise to different Raman bands. In the experiment, the total laser power at the sample is ~70 mW, so that the laser power at each focus is approximately 1 mW. It is obvious that a ~ 70 mW single focus totally destructs living cells. The distribution of the total laser power into multiple foci significantly reduces the photodamage. The overall measurement time is 20 sec and the images are reconstructed from 768 Raman spectra. This is equivalent to less than 25 msec for measuring one Raman spectrum. It is not possible to obtain a high S/N Raman spectrum in such a short exposure time by conventional Raman microspectroscopy.

Absolute quantitative imaging by Raman microspectroscopy

Raman signal intensity is strictly proportional to the number of molecules in the focal volume. The number of Raman scatted photons can be expressed as follows:

(Number of Raman scattered photons in unit time [/s])

= (Absolute Raman differential cross-section of molecule [cm2 / sr])

x (Photon flux density of the incident light [/s cm2]) x (Number of molecules in the focal volume)

x (Collecting solid angle [sr]) x (Detection efficiency [%]).

This equation indicates that the number/concentration of molecules in the focal volume can be estimated if the absolute Raman cross-section is known. The absolute Raman cross-section is a molecular property, which is a specific constant for a Raman band of a molecule and independent of experimental conditions except for the excitation wavelength. In order to determine absolute Raman cross-sections of biomolecules, polarized Raman measurements have been performed by using a confocal Raman microspectometer, whose polarization characteristics and the wavelength dependence of sensitivity are accurately calibrated. From the observed polarized Raman spectra, the parallel and perpendicular components of the absolute Raman cross-sections of several biomolecules are determined with accuracies of 20 %.

Figures 4 show the parallel and perpendicular Raman spectra of phenylalanine and cytochrome c reduced form (ferrous cyt. c) aqueous solutions. Phenylalanine has a marker band at 1003 cm(-1) and ferrous cyt. c at 603 cm(-1). From these polarized spectra, the parallel and perpendicular components of the absolute Raman cross-section of the 1003 cm(-1) Raman band are determined to be 0.96 and 0.05 x 10-29 [cm2/molecule sr], respectively. Those of the 603 cm(-1) band of ferrous cyt. c are determined to be 2.5 and 4.6 x 10-26 [cm2/molecule sr]. The absolute Raman cross-section of ferrous cyt. c is more than 1000 times larger than that of phenylalanine due to the resonance Raman effect, which enables sensitive and specific detection of ferrous cyt. c.

Using these values, the concentration of the phenylalanine residue in proteins and that of ferrous cyt. c are estimated in a cell (L929 NCTC) as shown in Figs. 5. The distribution of the phenylalanine residue, which is known as an indicator of protein abundance in a cell, is rather homogeneous inside the cell with 40~70 mM concentrations. On the other hand, that of ferrous cyt. c is highly localized at the marginal area of the cell, most probably reflecting the distribution of mitochondria in the cell. The estimated concentration of ferrous cyt. c in mitochondria is typically 10~30 μM. The concentration of phenylalanine in mitochondria is 50~70 mM, while that in the other area is about 40 mM. As far as the author is aware, these absolute numbers of concentrations have never been estimated with any other existing methods of molecular imaging.

New linear and nonlinear Raman microspectroscopic systems have been constructed. Raman imaging speed has been improved to be more than 10 times faster than that of the conventional confocal Raman microspectroscopy. Absolute quantitative imaging of a living cell is also demonstrated. The Raman cross-sections of the protein and ferrous cyt. c marker bands have been determined and their concentrations in a L929 NCTC cell estimated in situ without staining. A new potential of Raman microspectrosopy is thus shown for fast and absolutely quantitative molecular imaging

Fig. 1 (a) Typical CARS spectrum in the fingerprint region of a living budding yeast cell. (b) Im[X(3)] spectrum of (a) with MEM.

Fig. 2 Schematic of multi-focus confocal Raman microspectrometer

Fig. 3 Raman images of budding yeast cells of the Raman bands at (a) 1446 cm(-1), (b) 1583 cm(-1), (c) 1602 cm(-1), (d) 1655 cm(-1). The total acquisition time is approximately 20 sec. The scale bar is 4 μm.

Fig. 4 The parallel (I//) and perpendicular (I⊥) Raman spectra of phenylalanine (a) and ferrous cyt. c solutions (b).

Fig. 5 The molecular concentration images of a L929 (NCTC) cell. (a) phenylalanine, (b) ferrous cyt. c

本論文は、生細胞について、分子レベルで定量的情報を高速に取得するための新規顕微ラマン分光法の開発と、その応用について記述されており、全5章から構成される。

第1章では導入として、本研究の目的が、生細胞内分子の挙動を分子レベルで理解することにあり、そのためにバイオイメージングがいかに重要であるかが述べられている。そして、バイオイメージング手法として、既存の自発ラマン顕微分光装置およびコヒーレント・アンチ・ストークス・ラマン散乱(CARS)顕微分光装置が、細胞内分子の動的挙動を測定するための十分な高速性、定量性をもっていないことが説明されている。既存手法に代わる新規な手法として、定量的CARS顕微分光法、および自発ラマン散乱に基づいた、多焦点共焦点顕微ラマン分光法が提案されている。さらに、ラマン散乱の絶対定量性に着目し、生細胞中の生体分子数・濃度を見積もる、絶対定量的分子イメージングが提案されている。

第2章では、定量的CARS顕微分光法の詳細およびその生細胞への応用について述べられている。(サブ)ナノ秒パルス光を用いることで、生細胞の指紋領域の信号取得を可能とした。また、得られたCARSスペクトルに対して最大エントロピー法を適用することで、濃度に線形な非線形感受率の虚部スペクトルへと変換し、定量的な議論を可能にした。定量的CARS顕微分光法を生細胞に応用した結果が二つ述べられている。一つ目は、レーザー照射による出芽酵母細胞の死過程の研究である。本論文では、12 sec/イメージという高い時間分解能を達成することで、その動的挙動を追跡した。その結果、「生命のラマン分光指標」と呼ばれる1602 cm-1のラマンバンドが消失した後、「ダンシングボディ」と呼ばれる液胞由来の物質が出現することが示唆された。二つ目は、界面活性剤の生細胞の可溶化過程への応用である。細胞内分子と重水素化したドデシル硫酸ナトリウム(SDS)を同時に追跡した。本論文の結果から、SDSが数10 mMの濃度で細胞内の脂肪滴に選択的に蓄積していることが示唆された。さらに、時間分解CARSイメージとして、SDSが細胞内に徐々に取り込まれ、最終的に細胞が可溶化され、タンパク質が細胞外へと急激に流出する様子が可視化された。このように、細胞内分子の動的挙動を測定するためには、定量的CARS顕微分光法が非常に強力であることが示された。

第3章では、多焦点共焦点顕微ラマン分光法の詳細および性能評価、生細胞への応用が述べられている。マイクロレンズ・アレイ、ピンホール・アレイ、ファイバー・バンドルを用いることで、多焦点励起、共焦点配置、位置情報とスペクトル情報の同時取得を実現した。これにより、試料における48個の異なる焦点からの48本のラマンスペクトルの同時測定に成功した。面内0.3 μm, 奥行き1 μm 程度の空間分解能を有している。本装置により、16 x12 〓m2の領域における出芽酵母細胞のラマン分光イメージを20秒以下で取得可能とした。また、6分以下で3次元ラマン分光イメージを得ることが出来るようになり、出芽酵母細胞の細胞内構造を立体的に可視化した。このように、既存の単焦点共焦点顕微ラマン分光装置と比較して、数10倍の高速化に成功した。

第4章では、絶対定量的分子イメージングの試みが述べられている。生体分子の絶対ラマン散乱断面積を正確に決定し、分光計を厳密に較正することで、生細胞内の分子濃度を見積もった。L929(NCTC)細胞について、フェニルアラニン残基、還元型シトクロムc、リン酸の濃度イメージが、出芽酵母細胞について、フェニルアラニン残基、還元型シトクロムc、b、エルゴステロールの濃度イメージが得られている。これら複数の化学種の濃度イメージに基づいて、細胞小器官ごとの化学組成の定量的見積もりを可能とした。これは、他の手法では得られない、生細胞中分子の新たな定量的情報であり、顕微ラマン分光法による絶対定量的分子イメージングが、いかにユニークで強力な分析手法であるかを示している。

第5章は以上の研究成果のまとめおよび今後の展望である。

本研究により、高速定量的顕微ラマン分光法が開発され、これまで不可能であった、生細胞中の速い分子の動的挙動の測定、分子濃度に関する情報の取得が可能となった。本研究においてもさまざまな応用がなされているように、開発された手法が生細胞を研究する新しい方法論となることは明白である。このような新規の測定手法の開発とその有用性を提示した本論文の内容は高く評価できる。

本論文第2章の主要部分は、Angewandte Chemie International Edition 誌に公表済み(加納英明、Philippe Leproux、Vincent Couderc、James Day、Mischa Bonn、〓口宏夫との共著)、第3章の主要部はOptics Letters誌および応用物理学会誌に公表済み(ともに〓口宏夫との共著)であるが、論文提出者が主体となって実験および解析を行なっており、その寄与が十分であるので、学位論文の一部とすることに何ら問題はないと判断する。

以上の理由から、論文提出者奥野将成に博士(理学)の学位を授与することが適当であると認める。