Vibrational imaging of newly synthesized proteins in live cells by stimulated Raman scattering microscopy

Physical Principle of Isotope-Based SRS Imaging.

SRS microscopy is a molecular-contrast, highly sensitive imaging technique with intrinsic 3D sectioning capability. It selectively images the distribution of molecules that carry a given type of chemical bonds through resonating with the specific vibrational frequency of the targeted bonds (30, 33, 41). As Fig. 1A illustrates, by focusing both temporally and spatially overlapped Pump and Stokes laser pulse trains into samples, the rate of vibrational transition is greatly amplified by about 10⁷ times when the energy difference of the two laser beams matches the particular chemical bond vibration, Ω_vib (41). Accompanying such stimulated activation of one vibrational mode, one photon is created into the Stokes beam and simultaneously another photon is annihilated from the Pump beam, a process called stimulated Raman gain and stimulated Raman loss, respectively. Essentially, the energy difference between the Pump photon and the Stokes photon is used to excite the vibrational mode, fulfilling energy conservation. As shown in Fig. 1B, a high-frequency modulation scheme, where the intensity of the Stokes beam is turned on and off at 10 MHz, is used to achieve shot noise-limited detection sensitivity by suppressing laser intensity fluctuations occurring at low frequencies. The transmitted Pump beam after the sample is detected by a large-area photodiode, and the corresponding stimulated Raman loss signal, which also occurs at 10 MHz, is demodulated by a lock-in amplifier. By scanning across the sample with a laser-scanning microscope, a quantitative map with chemical contrast can be produced from the targeted vibrating chemical bonds. As the SRS signal is dependent on both Pump and Stokes laser beams, the nonlinear nature herein provides a 3D optical sectioning ability.

Here, we detect the vibrational signal of C–D as an indicator for newly synthesized proteins that metabolically incorporate deuterium-labeled amino acids (Fig. 1B). When hydrogen atoms are replaced by deuterium, the chemical and biological activities of biomolecules remain largely unmodified. Intriguingly, the C–D stretching motion displays a distinct vibrational frequency from all of the other vibrations of biological molecules inside live cells. It is known from classical mechanics that the frequency of vibrational oscillation, Ω_vib, inversely scales with the square root of the reduced mass of the oscillator , where k is the spring constant of the corresponding chemical bond, and μ denotes the reduced mass of the oscillator. The reduced mass of the C–D oscillator is increased by two folds when hydrogen is replaced by deuterium. Based on the above equation, Ω_vib would be reduced by a factor of . Indeed, the experimentally measured stretching frequency is shifted from ∼2,950 cm⁻¹ of C–H to ∼2,100 cm⁻¹ of C–D. Remarkably, the vibrational frequency of 2,100 cm⁻¹ is located in a cell-silent spectral window in which no other Raman peaks exist (Fig. S1), thus enabling detection of exogenous C–D with both high specificity and sensitivity.

SRS Imaging of Newly Synthesized Proteins by Metabolic Incorporation of Leucine-d₁₀ in Live HeLa Cells.

Among the 20 natural amino acids, leucine is an essential one with both high abundance in protein (∼9% in mammalian cells) and a large number of side-chain C–H that can be replaced by C–D (43). Hence, we first demonstrated the feasibility of our technique by detecting the metabolic incorporation of leucine-d₁₀ (l-leucine-2,3,3,4,5,5,5,5′,5′,5′-d₁₀ as shown in Fig. 2A) to nascent proteins in live HeLa cells. Fig. 2B shows the spontaneous Raman spectrum of HeLa cells incubated in the medium containing 0.8 mM free leucine-d₁₀ for 20 h (blue) overplotted with the spectrum of HeLa cells growing in the regular medium without leucine-d₁₀ (red) as well as the spectrum from a 10 mM free leucine-d₁₀ solution in PBS (black). As indicated by the comparison between the blue and the red spectra, the Raman peaks of leucine-d₁₀, exhibiting multiple peaks around 2,100 cm⁻¹ due to symmetric and asymmetrical C–D stretching, are indeed located in the cell-silent region. The comparison of the blue and the black spectra implies that leucine-d₁₀ incorporated into cellular proteome after 20 h is enriched to about 10 mM. Thus, a 10% incorporation yield of leucine-d₁₀ can be estimated at this condition based on the intrinsic leucine concentration of about 100 mM in proteins (calculated from protein concentration and leucine percentage in cells).

Based on the above spectra, we choose to target the central 2,133 cm⁻¹ vibrational peak of C–D to acquire SRS images of nascent proteins in live HeLa cells. As expected, HeLa cells growing in regular medium show no detectable SRS contrast at 2,133 cm⁻¹ (Fig. 2C), which is consistent with the flat spectral baseline (red in Fig. 2B) in the cell-silent region. In contrast, SRS image of HeLa cells growing in the medium containing 0.8 mM leucine-d₁₀ (Fig. 2D) shows a weak but clearly identifiable contrast outlining the cell shape. As a control, the off-resonant SRS image at 2,000 cm⁻¹ of the same cells is background free (Fig. 2E). Such clean chemical contrast among Fig. 2 C–E would be difficult for CARS microscopy due to the presence of its nonresonant background. As a protein reference, an image taken at 2,940 cm⁻¹ [CH₃ stretching mainly from proteins with minor cross talk from lipids (33)] shows both existing and newly synthesized proteins (Fig. 2F), the signal of which comes from the same regions but is much stronger than that in Fig. 2D. Thus, we have demonstrated the feasibility of using SRS imaging to detect newly synthesized proteins by specifically targeting the C–D vibrational signal of metabolically incorporated leucine-d₁₀ in live HeLa cells. This opens up an imaging opportunity to capture nascent proteome dynamics in live cells under a myriad of cues.

Imaging Optimization by Metabolic Incorporation of a Deuterium-Labeled Set of All Amino Acids in Live HeLa Cells with Multicolor SRS Imaging.

Although leucine is the most abundant essential amino acid, it only accounts for a small fraction of amino acids in proteins. Hence, we reasoned that deuterium labeling of all of the amino acids would lead to a substantial signal enhancement. Indeed, the spontaneous Raman spectrum (Fig. 3A) of HeLa cells incubated with a deuterium-labeled set of all 20 amino acids (prepared by supplying a uniformly deuterium-labeled whole set of amino acids to leucine-, lysine-, and arginine-deficient DMEM; for more details, refer to Materials and Methods) exhibits C–D vibrational peaks about five times higher than the blue spectrum in Fig. 2B under the same condition. The corresponding SRS image at 2,133 cm⁻¹ (Fig. 3B) shows a significantly more pronounced signal than that in Fig. 2D under the same intensity scale. In particular, nucleoli (indicated by arrows in Fig. 3B and verified by differential interference contrast visualization) exhibit the highest signal, which is in accordance with previous reports using BONCAT and our own fluorescence staining results (Fig. S2). Nucleoli, the active sites for ribosomal biogenesis, have been reported to involve rapid nucleolar assembly and proteomic exchange (44–46). Such fast protein turnover is indeed reflected by the spatial enrichment of newly synthesized protein signals in those subcellular areas (Fig. 3B). Note that SRS imaging here is directly performed on live cells and hence free from potential complications due to fixation and dye conjugation. Again, the off-resonant image at 2,000 cm⁻¹ is clean and dark (Fig. 3C), proving the specificity of SRS imaging of C–D at 2,133 cm⁻¹. In addition to imaging newly synthesized proteins, SRS can readily image intrinsic biomolecules in a label-free manner. By simply adjusting the energy difference between the Pump and the Stokes beams to match the vibrational frequency of amide I, lipids, and total proteins, respectively, Fig. 3 D–F shows the SRS images of amide I band at 1,655 cm⁻¹ primarily attributed to proteins; CH₂ stretching at 2,845 cm⁻¹ predominantly for lipids; and CH₃ stretching at 2,940 cm⁻¹ mainly from proteins with minor contribution from lipids.

Time-Dependent de Novo Protein Synthesis and Protein Synthesis Inhibition.

Being linearly dependent on analyte concentration, SRS contrast is well suited for quantification of de novo protein synthesis in live cells. Here, we show time-dependent protein synthesis images under the same intensity scale (Fig. 4 A–C). As expected, the new protein signal (2,133 cm⁻¹) from 5-, 12-, and 20-h incubation increases substantially over time (Fig. 4 A–C), whereas the amide I (1,655 cm⁻¹) signal remains at a steady state (Fig. 4 D–F). Because protein distribution is often heterogeneous in biological systems, we presented a more quantitative representation by acquiring ratio images between the newly synthesized proteins and the total proteome (from either amide I or CH₃). Fig. 4 G–I depicts the fraction of newly synthesized proteins (2,133 cm⁻¹) among the total proteome (1,655 cm⁻¹) and its spatial distribution. The fraction of newly synthesized proteins is growing with time from 5 to 20 h, gradually highlighting nucleoli as the subcellular compartments with fast protein turnover (44–46). Such quantitative ratio imaging of new versus old proteomes would be very difficult to obtain using BONCAT or mass spectroscopy without the destruction of cells. More time-dependent cell images are shown in Fig. S3. Moreover, Fig. 4J shows time-lapse SRS images of a live dividing HeLa cell after 20-h incubation in deuterium-labeled, all-amino acids medium, clearly proving the viability of cells under the imaging condition.

The effect of protein synthesis inhibition by chemical drugs is further tested to validate that the detected C–D signal indeed derives from nascent proteins. HeLa cells incubated with a deuterium-labeled set of all amino acids together with 5 μM anisomycin, which works as a protein synthesis inhibitor by inhibiting peptidyl transferase or the 80S ribosome system, for 12 h show the absence of the C–D signal in the spontaneous Raman spectrum (Fig. 4K). Furthermore, SRS imaging of the same samples (Fig. 4L) exhibits drastically weaker signal [Fig. S4 provides a more thorough analysis of the residual signal, which is possibly attributed to the intracellular free amino acid pool (47)] compared with Fig. 4B without the protein synthesis inhibitor. As a control, the corresponding 2,940 cm⁻¹ image (Fig. 4M) of total proteome remains at a similar level as the non–drug-treated counterpart in Fig. 3F. Thus, the detected C–D SRS signal (Fig. 4 A–C) originates from deuterium-labeled nascent proteins, which vanishes upon adding the protein synthesis inhibitor.

Demonstration on HEK293T Cells and Neuron-Like Differentiable Neuroblastoma N2A Cells.

To show the general applicability and potential of our method, we choose two additional mammalian cell lines for further demonstration: human embryonic kidney HEK293T cells, and neuron-like neuroblastoma mouse N2A cells, which can be induced to differentiate with the growth of neurites (i.e., axons and dendrites). The spontaneous Raman spectrum (Fig. 5A) of HEK293T cells incubated with a deuterium-labeled set of all amino acids for 12 h exhibits a 2,133 cm⁻¹ C–D channel signal nearly as high as the 1,655 cm⁻¹ amide channel signal. The resulting SRS image shows a bright signal for new proteins with an intense pattern residing in nucleoli (Fig. 5B). As before, the off-resonant image (2,000 cm⁻¹) displays vanishing background (Fig. 5C); the amide I channel (1,655 cm⁻¹) image (Fig. 5D) exhibits consistent overall proteome distributions similar to that in HeLa cells; CH₂ channel (2,845 cm⁻¹) image (Fig. 5E) depicts a more diffusive lipid distribution in cytoplasm compared with that in HeLa cells. Consistent with the results obtained in HeLa cells above, the ratio image (Fig. 5F) between the newly synthesized proteins (Fig. 5B) and the total proteins (Fig. 5D) highlights nucleoli for active protein turnover in HEK293T cells as well (44–46).

In addition to showing the ability to image newly synthesized proteins inside cell body, our technique can also be applied to tackle more complex problems, such as de novo protein synthesis in neuronal systems (2–4). As an initial demonstration, we imaged the newly synthesized proteins in neuron-like N2A cells, which have been extensively used as a model system to study neuronal differentiation, axonal growth, and signaling pathways. Under differentiation condition, N2A cells massively grow new neurites from cell bodies and form connections with other cells. Fig. 6A shows the image of newly synthesized proteins after induction for differentiation, by simultaneously differentiating the N2A cells and supplying with the deuterium-labeled set of all amino acids for 24 h. Similar to HeLa and HEK293T cells, N2A cell bodies are observed to display high-level protein synthesis. More interestingly, newly synthesized proteins are also observed in a subset of, but not all neurites (Fig. 6 A and B), which implies that the observed neurites in Fig. 6A are newly grown under the differentiation condition. For a detailed visualization, Fig. 6 C and D shows the zoomed-in regions in the dashed squares in Fig. 6 A and B, respectively. A more comprehensive examination is illustrated by both the ratio image (Fig. 6E) between Fig. 6 C and D and the merged image (Fig. 6F) with the red channel designating new protein signal from Fig. 6C and the green channel designating total protein signal from Fig. 6D. On one hand, both the ratio image and the merged image highlight the neurites with higher percentage of new proteins (indicated by stars), implying these neurites are newly grown. On the other hand, from the merged image, there are some neurites (indicated by arrows) showing obvious signals in the green channel (total proteins) only but with no detectable signal in the red channel (new proteins). Hence, the arrow indicated neurites are most likely older than their starred counterparts. In addition, the transition from green to red in the merged image (Fig. 6F) implies the growth direction by which new neurites form and grow. A second set of N2A images showing similar patterns as in Fig. 6 is also examined in Fig. S5. A more relevant system to study de novo protein synthesis and neuronal activities would be hippocampal neurons, which are known to be involved in long-term memory formation (2–4). SRS image (2,133 cm⁻¹) of hippocampal neuron cells incubated with a deuterium-labeled set of all amino acids shows a newly synthesized protein pattern in the neurites (Fig. S6). The intricate relationship between protein synthesis and neuronal activities is currently under investigation.

SRS Microscopy.

The experimental setup is shown in Fig. 1B. Spatially and temporally overlapped pulsed Pump (tunable from 720 to 990 nm, 7 ps, 80-MHz repetition rate) and Stokes (1,064 nm, 5∼6 ps, 80-MHz repetition rate, modulated at 10 MHz) beams, which are provided by picoEMERALD from Applied Physics & Electronics are coupled into an inverted laser-scanning microscope (FV1000 MPE; Olympus) optimized for near-IR throughput. A 60× water objective (UPlanAPO/IR; 1.2 N.A.; Olympus) is used for all cell imaging. After passing through the sample, the forward going Pump and Stokes beams are collected in transmission by a high N.A. condenser and imaged onto a large area Si photodiode. A high OD bandpass filter (890/220, Chroma) is used to block the Stokes beam completely and to transmit the Pump beam only for the detection of the stimulated Raman loss signal. The output current from the photodiode is terminated, filtered, and demodulated by a lock-in amplifier (SR844; Stanford Research Systems) at 10 MHz to ensure shot noise-limited detection sensitivity. For imaging, 512 × 512 pixels are acquired for one frame (26 s per frame) with a 100-μs pixel dwell time and 20-μs time constant from the lock-in amplifier. Powers after 60× IR objective used for imaging are as follows: 61 mW for modulated Stokes beam; 145 mW for the Pump beam of 2,133 cm⁻¹, 2,000 cm⁻¹, and 1,655 cm⁻¹ channels; and 64 mW for Pump beam of 2,950 cm⁻¹ and 2,845 cm⁻¹ channels.

Metabolical Labeling of the Newly Synthesized Proteins by Deuterium-Labeled Amino Acids.

Deuterium-labeled leucine-d₁₀ medium is made by adding leucine-d₁₀ (0.8 mM), lysine (0.8 mM), and arginine (0.4 mM) (Sigma) into leucine-, lysine-, and arginine-deficient DMEM (Sigma). Deuterium-labeled all-amino acids medium is made by adding uniformly deuterium-labeled amino acid mix (20 aa) (Cambridge Isotope) into leucine-, lysine-, and arginine-deficient DMEM (Sigma). The final concentration of leucine-d₁₀ is adjusted to be 0.8 mM among the amino acid mix. (Because the starting medium is leucine, lysine, and arginine deficient, by adding the deuterium-labeled 20-aa mix, we essentially deuterate all of the leucine, lysine, and arginine as well as about one-half of the other amino acids.) Cells are seeded on a coverslip in a petri dish with 2 mL of regular DMEM with 10% (vol/vol) FBS and 1% penicillin/streptomycin (Invitrogen) for 20 h. The regular medium is then replaced with medium containing either leucine-d₁₀ or a deuterium-labeled set of all amino acids. After incubation for a certain amount of time, the coverslip is taken out to make an imaging chamber filled with PBS for SRS imaging. For N2A cells, in the process of induced cell differentiation with serum deprivation and 1 μM retinoic acid, the deuterium-labeled set of all amino acids is supplemented.

Spontaneous Raman Spectroscopy.

The spontaneous Raman spectra were acquired using a laser Raman spectrometer (inVia Raman microscope; Renishaw) at room temperature. A 27-mW (after objective), 532-nm diode laser was used to excite the sample through a 50×, N.A. 0.75 objective (NPLAN EPI; Leica). The total data acquisition was performed during 80 s using the WiRE software.