11. Conclusion

In summary, we have discovered leucine pools in a nascent E. coli biofilm by using Raman imaging, where high levels of leucine were accumulated. We have also demonstrated that Raman imaging is a powerful tool for detecting in situ and identifying key biomolecules involved in biofilms even if they are yet unknown (i.e., possibility of exploratory research). We hope this study will open up new avenues for developing a simple but effective chemical means that will enable molecular-level elucidation of bacterial biofilms.

33

Chapter IV

Visualization of Dynamic Proteome Localization to Lipid Droplets in

Single Fission Yeast Cells Using

Stable Isotope Probing

34

IV-1. Introduction

Lipid droplets (LDs), also called lipid bodies or fat bodies, are globular organelles that are ubiquitously found in most eukaryotic cells from yeast to mammals. LDs are cytoplasmic structures which store neutral lipids in its core. Unlike other intra cellular organelles (aqueous compartments separated by lipid bilayers), LDs have a unique physical structure. LDs’ core is surrounded by a single layer of phospholipids (polar) with some embedded proteins, separating it from the aqueous cytoplasmic space. The core is primarily constituted by triacylglycerols (TAG) and steryl esters (SE), crucial substances for cells (Figure IV-1). These molecules cannot serve as direct building blocks for lipid bilayers as they lack charged groups. Instead, upon requirement, these can be hydrolysed to sterols, diacylglycerol and fatty acid which can mainly be used to derive energy.

LDs were long considered as a static energy storage units and thus remained the least characterized cytoplasmic organelle. In the past half-decade, however, LDs have come under the spotlight of cell biology. There is now growing evidence that reveals diverse roles and an intrinsically dynamic nature of LDs as a key player in various cellular processes, such as lipid

Figure IV-1. Schematic representation of a lipid droplet.

Lipid core (blue) primarily contains TAG and SE which is surrounded by a phospholipid monolayer (purple).

35

homeostasis and cell signalling [36]-[41]. It has also been shown that LDs are relevant to many serious health issues, including obesity, type 2 diabetes, and atherosclerosis. Among various functions of LDs, a recently proposed hypothesis that LDs serve to temporarily sequester proteins is thought provoking and might potentially deepen our understanding of LD [42]-[44].

Proteomic studies have detected LD-associated proteins in many cell types [43], [45], which apparently have little to do with lipid metabolism. Those proteins include histones [42], caveolins [46], and perilipin family proteins [47]. Although proteomic analysis is very powerful for identifying and characterizing individual key proteins, biochemical fractionation and purification procedures adopted in LD proteomics are vulnerable to contamination. Moreover, proteomic analysis cannot provide information on spatial localization (distribution) of proteins as well as their temporal evolution. To test whether LDs are both spatiotemporally and functionally associated with proteins (e.g., sequestration), direct evidence needs to be obtained from single living cells using microscopic and imaging techniques.

IV-2. How can Raman microspectroscopy help?

Molecular imaging based on linear/nonlinear Raman spectroscopy has emerged as a promising tool to trace intracellular processes in vivo and at the molecular level. In contrast with commonly employed fluorescence microscopy, Raman-based methods require no exogenous probe to be introduced to cells. Vibrational resonances, which are an inherent property of molecules, give rise to chemical specificity and, hence, enable label-free molecular imaging. Of particular importance is work using coherent anti-Stokes Raman scattering (CARS) microscopy, which has high sensitivity and three-dimensional sectioning capability.

The Xie group at Harvard demonstrated video-rate CARS imaging of living cells and tissues [48]. Hellerer et al. [49] also used CARS microscopy to visualize lipid distributions in Caenorhabditis elegans. Very recently, Xie and coworkers have further improved image contrast

36

by using stimulated Raman scattering (SRS) instead of CARS [50], [51]. SRS is, in principle, free from the nonresonant background that always interferes with vibrationally resonant signals in CARS. These nonlinear Raman microscopic studies achieved to date mostly rely on the strong C-H stretch vibrations around 2850 cm^-1 [48]-[53]. However, C-H stretch images alone do not allow us to directly look into the interplay between LDs and proteins and other organelles. In contrast, Raman microspectroscopy and imaging [28], [54]-[57], although taking longer data acquisition time, provide more extensive and detailed molecular information than CARS microscopy without the need for complicated spectral analysis [58]. Hence we use Raman microspectroscopy and imaging.

IV-3. Stable isotope probing

Since we want to learn spatio-temporal and functional relationships between proteome and lipidome, it is essential to be able to distinguish the biomolecular components that are being actively synthesized from those that already exist. Such a distinction is not possible by ordinary Raman spectroscopy. To achieve this capability, we make use of stable isotopes since isotopic substitution can be of invaluable help in the analysis of vibrational spectra.

Stable-isotope probing (SIP), in which stable isotopes such as ¹³C and ¹⁵N are incorporated in cells as a non-perturbative tracer for RNA [59], DNA [60], and proteins [61], has been widely used for microbial identification. When combined with Raman imaging [8], [62]-[64], the stable-isotope labeling strategy can confer an ability to differentiate between cellular components produced through distinct anabolic pathways and temporal evolutions.

This ability arises from the fact that the characteristic frequency (ν) of an oscillator representing a molecular vibration is inversely proportional to the square root of the reduced mass (µ) of the oscillator (Equation IV-1).

37

𝑣 =_2𝜋¹ √^𝑘_µ . . . 𝐄𝐪𝐮𝐚𝐭𝐢𝐨𝐧 𝐈𝐕 − 𝟏

In general the force constants (k) are unaltered by isotopic substitution. Hence, the heavier the incorporated isotope, the lower the frequency i.e., redshift in a Raman spectrum (Figure IV-2)[65].

IV-4. Sample

Raman microspectroscopy is so powerful in that it can be applied to various samples starting from single cells [8], [28], [30], biofilms [3], [27], tissues [48], [66], [67] to small living animals such as mice [68]-[70] and humans [71]. Since our main interest is to study proteome and lipidome together, it is important to study a whole organism. So, in the present study, we chose fission yeast, Schizosaccharomyces pombe (S. pombe), which is one of the most popular eukaryotic model organisms in molecular and cellular biology, owing to its size. Moreover, LD

Normal isotope

heavy isotope

Figure IV-2. Illustration of isotope-induced red shift in a Raman spectrum. The red trace corresponds to a normal isotope and blue to its heavier counterpart.

38

biology of yeasts has been extensively studied and a wealth of information is already available.

Main reasons behind choosing fission yeast are summarized as follows:

a) It is a unicellular rod-shaped eukaryote

b) Typical size ranges from 3 - 5 µm in diameter and 8 - 15 µm in length

c) They grow and divide by medial fission forming two equally sized daughter cells which is very suitable for cell cycle research and

d) Yeasts are shown to form/accumulate lipid droplets at various conditions

Figure IV-3. Scanning electron micrograph of a fission yeast (S.pombe) culture.

Scale bar=10µm.

Source: The cell cycle. Principles of control. David O Morgan (Jan 1^st 2007)

39

IV-5. Growth of yeast cells in

¹³

C medium

As a first step, to examine whether there are any adverse effects on cell growth caused by heavy carbon isotope (¹³C), we cultured fission yeast cells in minimal medium with ¹³ C-glucose supplied as a primary carbon source and characterized the growth curve. Briefly, S.

pombe cells were cultured in both ¹²C- and ¹³C-glucose containing Edinburgh Minimal Medium (EMM) broth and the increase of optical densities (OD) with time, taken at 600 nm after 10 times dilution of the culture to avoid saturation at longer time, using a UV-Vis spectrophotometer gives a measure of density of both the cultures. Both cultures showed quite a similar growth characteristics (Figure IV-4). The composition of EMM broth is tabulated in Table IV-1. Comparison of growth curves in Figure IV-4 indicates that cells grown in ¹³ C-EMM perform as normal metabolic activities as in ¹²C-EMM. Thus we can safely say that the use of ¹³C-glucose instead of ¹²C-glucose does not affect the growth of yeast cells.

Figure IV-4. Growth curves of fission yeast in¹²C- and ¹³C-glucose containing Edinburgh minimal medium.

40

IV-6. Stable isotope labeling of fission yeasts

After confirming that ¹³C is harmless, space-resolved Raman spectra from yeast cells grown in ¹²C- and ¹³C-glucose medium were recorded. In ¹²C medium, Raman spectra obtained from cytoplasmic regions are dominated by vibrational bands associated with cellular proteins, [hereafter called protein rich (PR) spectra], which primarily include the backbone amide modes and amino acid side chain modes. From Figure IV-5 (b), it is clear that yeast cells grown in EMM with regular ¹²C-glucose have the amide I mode of proteins at 1654 cm^–1, amide III at 1253cm^–1, C-H bending modes at 1451 and 1338 cm^–1, aromatic ring breathing mode of phenylalanine (Phe) residues at 1003 cm^–1 and a tyrosine band at 853 cm^–1. Bands between 1040 - 1150 cm^–1 and at 916 cm^–1 represent the C-C stretching modes. When cells are fed with stable ¹³C isotope containing EMM, several bands show red shift due to heavy isotope effect.

Primarily, the amide I and III shift to 1620 and 1237cm^–1, respectively. The ring breathing mode of Phe shifts from 1003 to 967 cm^–1 and strong tyrosine band appears at 828 cm^–1. The C-C stretch modes also show small shifts including the relatively isolated 916 cm^–1 band, which shifts to 899 cm^–1.

Table IV-1. Composition of Edinburgh Minimal Medium.

41

Figure IV-5. (a) Optical image of a typical fission yeast. The transparent region inside the cell indicated by an asterisk is the cytoplasm where protein rich spectra were recorded. The black dot, highlighted by a shown in red square represents LD where lipid rich spectra were measured. Scale bar = 5µm. (b,c) Space-resolved Raman spectra from yeast grown in 12C and 13C-EMM in protein rich (b) and lipid rich regions (c) to identify vibrational isotope shifts.

(b)

(c)

(a)

42

Similarly, Raman spectra [called lipid rich (LR) spectra] obtained from LDs are dominated by the C=C stretch of unsaturated lipid chains, C-H bending modes of the hydrocarbon chains, and skeletal C-C stretch modes as shown in Figure IV-5 (c). Obvious red shifts can be observed for the C=C stretch from 1654 to 1595 cm^-1 and the ergosterol marker band (arising from the C=C diene in-phase stretch of the ring moiety) at 1602 to 1542 cm^-1 in

12C and ¹³C media, respectively. The C=O stretch of the ester linkage shifts from 1744 to 1695 cm^-1. A band at 716 cm^–1 ascribed to the phospholipid head group shifts to 697 cm^–1. The C-H bending modes, including CH2 scissors and CH3 degenerate deformation of the hydrocarbon chains, at 1440/1450 and 1338 cm^-1 show no appreciable isotope shifts. The in-phase CH2

twisting mode at 1301 cm^–1 is observed at the same frequency in both media and is completely insensitive to ¹³C substitution. All vibrational band shifts observed in the PR and LR spectra are consistent with previous reports [8], [28], [62], [72]. Details of peak position, assignment and ¹³C isotope effect are tabulated for proteins and lipids in Table IV-2.

43

IV-7. Selection of marker bands for proteins and lipids

Among all bands in PR spectra, the ¹³C-shifted Phe bands at 967 cm^–1 appears in a clean spectral region and represents a strong, distinguishable Raman signal that can be easily detected even in cellular Raman spectra. Thus this band emerges as a potential marker for the ¹³C incorporation/substitution in newly formed proteins. For similar reasons, the band at 1542 cm^–1 in LR spectra helps as a marker for the newly synthesized lipids. Additionally, the lipid band at 1301 cm^–1, which does not show any shift, can be used as isotope insensitive lipid marker to see total intracellular lipids.

Table IV-2. Peak positions in ¹²C- and ¹³C-EMM, molecular component with their tentative assignments

44

The reason behind not considering C-C stretch modes as a possible marker is due primarily to the fact that these bands are observed in both PR and LR spectra and also because they are present in a highly congested region containing several bands, which makes them complicated for further analysis. Typically, tyrosine residues show a doublet at 853 and 825 cm^–1, which arises from the Fermi resonance between the ring breathing vibration and the overtone of an out-of-plane ring-bending vibration. This tyrosine doublet is known to be a strong indicator of the H-bonding strength of the phenolic hydroxyl group [26]. The tyrosine band at 853 cm^–1 (apparently shifting to 828 cm^–1 in ¹³C medium) was not considered as a potential protein marker because it is ambiguous whether the band intensity increase of the 828 cm^–1 band is really due to ¹³C isotope substitution or due to change in the H-bonding environment.

IV-8. Bulk experiments

Detailed space- and time-resolved Raman experiments were performed in order to understand the dynamics of ¹³C incorporation/substitution in lipids and proteins and to observe intermediary spectral features, if any, during the process. Raman spectra characteristic of PR and LR regions that can be free from cell individuality can be best represented by average spectra obtained from measurements on many cells. In light of this, we recorded Raman spectra in both PR and LR regions of 25 randomly picked yeast cells at each time (after washing twice with sterile distilled water) and obtained their averaged spectra as shown in Figures IV-6 (a) and IV-7 (a). Pearson correlation coefficients r, calculated using the formula below, for the

whole spectrum obtained are very high with

similarity greater than 90% as shown in Figures IV-6 (b) and IV-7 (b).

45

Figure IV-6. (a) Space- and time-resolved Raman spectra showing the evolution of isotope incorporation in PR regions in yeasts, (b) Percentage similarity among PR spectra at each time calculated by Pearson correlation coefficients.

Each spectrum is an average of 25 different cells.

(a)

(b)

46

Figure IV-7. (a) Space- and time-resolved Raman spectra showing the evolution of isotope incorporation in LR regions in yeasts, different cells, (b) Percentage similarity among LR spectra at each time calculated by Pearson correlation coefficients.

Each spectrum is an average of 25different cells.

(a)

(b)

47

This result shows that the variation between the measurements is small at each time and indeed individual spectra are very close to the averaged one. Figure IV-8 displays the time evolution of the intensities of several pairs of Raman bands observed in PR and LR spectra. To ascertain the degree of ¹³C substitution, ring breathing modes of Phe residues (1003 cm^–1 in ¹²C& 967 cm^–1 in ¹³C) and appearance of 1542 cm^–1 are used as markers for proteins and lipids, respectively. In general, the doubling time of fission yeasts in EMM broth is somewhere between 2 - 4 hr. Consistent with this time scale, at 3.5 hr, just after 1 generation, the intensities of ¹²C-amide I band [1654 cm^–1, Figure IV-8 (a)] and ¹²C-Phe band [1004 cm^–1, Figure IV-8 (b)] in PR spectra decrease considerably and at the same time ¹³C-Phe band at 967 cm^–1 becomes identifiable, though very weak, indicating active protein synthesis.

Approximately after a couple of generations, (~ 8 hr), appreciable amounts of ¹³C incorporation can be observed by the presence of equally strong 967 cm^–1 [see Figure IV-6 (a)].

By this time, the proteins in cells are made predominantly from ¹³C substrates as can be seen from the appearance of the ¹³C-shifted amide I backbone mode of proteins peaking at 1620 cm^–

1 [Figure IV-8 (a)]. ¹²C content in proteins fades away rapidly with time and eventually becomes negligible in just about 15 hr as indicated by the 1654 cm^–1 band in Figure IV-8 (a)

& (b). Simultaneously, incorporation of ¹³C in cytoplasmic proteins is >75% within that period of time and a complete incorporation can be seen within a day.

Similarly, in LR spectra, the ¹³C marker band at 1542 cm^–1 (shifting from 1602 cm^–1) band becomes identifiable in about 3.5 hr but becomes more obvious only after ~ 8 hr [Figure IV-8 (c)]. In order to understand the dynamics of this band, it is also important to know the fate of its complementary band at 1602 cm^–1. Interpreting the dynamics of 1602 cm^–1 band just by looking at the time-resolved spectra is complicated because the C=C stretch band at 1654 cm^–1 also shifts to 1595 cm^–1 and interferes with the observation of the 1542 cm^–1 band. As a result,

48

the intensity at 1602 cm^-1 decreases initially, reflecting the actual decrease of the band, but starts increasing from 15 hr because ¹³C-substituted C=C stretch band comes in [Figure IV-8 (d)].

Hence, it is only appropriate to correlate the appearance of the1542 cm^–1 band in an indirect manner,with the decay of the 1654 cm^–1 instead of the 1602 cm^–1 band. This is why the dynamics of the 1654 and 1542 cm^–1 bands are coupled together in Figure IV-8 (c). The intensity of the 1542cm^–1 band increases with time, whereas that of 1654 cm^–1 decreases. In lipids, ¹²C content becomes negligible at 23 hr and incorporation of ¹³C completes at around 30 hr. By repeating these experiment and looking into the dynamics of the 1654 cm^–1 band in proteins and lipids, we understand that complete depletion of the ¹²C-associated bands takes

<15 hr for proteins and ~24 hr for lipids. In other words, isotope incorporation in proteins occurs at a faster rate when compared to lipids.

49

(a) (b)

(c) (d)

Figure IV-8. ¹³C incorporation dynamics observed with incubation time in bulk experiments. (a) Amide I bands in PR spectrum at 1654 cm^–1 (red) and 1620 cm^–1 (blue) (b) Phenylalanine breathing mode in PR spectrum at 1003 cm^–1 (red) and 967 cm^–1 (blue) (c) ¹²C=¹²C str at 1654 cm^–1 (red) and ¹³C-shifted 1602 at 1542 cm^–1 (blue) from LR spectrum (d) 1602 cm^-1 from LR spectrum

50

IV-9. Single cell Raman imaging

We now have gained a detailed picture of how Raman spectra of the cytoplasm and LD vary as ¹³C-glucose is assimilated in S. pombe cells. However, those data were obtained by averaging behaviours of many different S. pombe cells, so they do not provide information on spatial distributions of the cellular components within a single cell and their temporal changes during the course of ¹³C incorporation. To address these crucial issues, we need to combine the

13C labeling technique with time-lapse Raman imaging on a single S. pombe cell.

Here, we use the Phe band to generate protein Raman images since Phe residues are usually abundant in proteins. The Phe band can probe nearly the whole intracellular population of proteins at a given instant. In other words, this Raman band manifests the proteome.

For markers of LDs, we use the bands at 1301 and 1602 cm^–1. Figure IV-9 shows time-lapse Raman images over 37 hr, constructed at 1003, 967, 1301, and 1602 cm^–1, of a single stationary-phase S. pombe cell with green fluorescence protein (GFP)-labeled mitochondria fed with ¹³C-glucose. All the four Raman images at a given measurement time were obtained simultaneously from one two-dimensional scan of the cell. Each Raman image is represented in a pseudocolor scale (red, magenta, cyan, yellow, or rainbow). For example, a red pseudocolor scale in which the highest intensities appear white, the moderate appear red, and the lowest appear black is used for the Raman images at 1003 cm^–1 [Figure IV-9 (c)].

Also shown are bright-field optical micrographs [Figure IV-9 (a)] and GFP fluorescence images [Figure IV-9 (b)] of the S. pombe cell. The optical micrographs are silent about molecular distribution, although LDs of ~0.1 mm size can be seen as black dots (as indicated by arrows at 31 hr). The GFP image clearly visualizes mitochondrial distribution at each time.

51

Let us examine first a pair of the time-lapse Raman images of the Phe band for proteins.

The Raman images at 1003 cm^–1 [Figure IV-9 (c)] represent the distribution of Phe-containing proteins that are originally present within the S. pombe cell when we started the imaging experiment, whereas those at 967 cm^–1 [Figure IV-9 (d)] map the distribution of newly formed proteins using exogenous ¹³C substrate. One hour after inoculation in ¹³C-EMM, proteins are

Figure IV-9. Time-lapse multimode Raman imaging of a single living S.pombe cell grown ¹³C-glucose containing medium.

(a) Bright-field optical images of the target cell. Arrows at 31 hr indicate the regions where there are many lipid droplets. LDs are identified as black dots. (b) GFP fluorescence images of mitochondria of the cell. (c–f) Raman images constructed at 1003 (Phe breathing mode of 12C-proteins), 967 (Phe breathing mode of 13C-substituted proteins), 1301 (in-plane CH2 twist of lipids), and 1602 cm-1. The scale bar in (a) measures 5 µm and also applies to the other images. (g) Correlation images between (d) and (e). The correlation coefficient at each pixel was computed using the equation given in the text

52

distributed homogenously across the cell and ¹³C is not yet incorporated into proteins. As time progresses, the ¹²C content in proteins is depleted particularly in two regions where many LDs are seen. After 31 hr, the intensity of the 1003 cm^-1 band decreases to a noise level. Protein synthesis machinery, which is known to be continuously operating, utilizes ¹³C substrate taken up in the cell to fulfill its mission. As a result, the concentration of ¹³C-labeled proteins probed at 967 cm^-1 gradually increases [Figure IV-9 (d)], which is consistent with the results of the bulk experiments. In sharp contrast with the existing ¹²C proteins, newly formed ¹³C-labeled

distributed homogenously across the cell and ¹³C is not yet incorporated into proteins. As time progresses, the ¹²C content in proteins is depleted particularly in two regions where many LDs are seen. After 31 hr, the intensity of the 1003 cm^-1 band decreases to a noise level. Protein synthesis machinery, which is known to be continuously operating, utilizes ¹³C substrate taken up in the cell to fulfill its mission. As a result, the concentration of ¹³C-labeled proteins probed at 967 cm^-1 gradually increases [Figure IV-9 (d)], which is consistent with the results of the bulk experiments. In sharp contrast with the existing ¹²C proteins, newly formed ¹³C-labeled

在文檔中 利用顯微拉微光譜和拉曼成像法研究大腸桿菌生物膜和利用同位素標記法研究活體裂殖酵母菌細胞代謝動力學 (頁 42-0)