Raman spectroscopy as a tool for detecting ... - Wiley Online Library

Research article Received: 29 October 2015

Revised: 4 March 2016

Accepted: 10 March 2016

Published online in Wiley Online Library: 19 April 2016

(wileyonlinelibrary.com) DOI 10.1002/jrs.4930

Raman spectroscopy as a tool for detecting mitochondrial fitness Nika Erjavec,a Giulietta Pinatoa and Kerstin Ramserb* Raman spectroscopy allows the molecular chemical analysis of whole living cells by comparing them to known Raman signatures of specific vibrational bonds. In this work we used Raman spectroscopy to differentiate between wild type yeast cells and mutants characterized by increased or reduced mitochondrial fragmentation. To associate mitochondrial fragmentation with biochemical markers, we performed Linear Discriminant Analysis (LDA) of whole cell Raman spectra (~50–100 cells/spectrum). We show that the long-lived, less fragmented mutants fall into a significantly distant cluster from the wild type and short-lived, more fragmented mutants. Clustering depends on respiratory growth and coincides with that of membrane phospholipids and some respiratory chain components. Spectral clustering is supported by enzymatic activity measurements of OXPHOS Complexes. In addition, we find that NAD(P)H autofluorescence also correlates with mitochondrial fragmentation, representing another likely aging biomarker, besides phospholipids and OXPHOS components. In summary, we demonstrate that Raman spectroscopy has the potential to become a powerful tool for differentiating healthy from unhealthy aged tissues, as well as for the prognostic evaluation of mitochondrial function and fitness. © 2016 The Authors Journal of Raman Spectroscopy Published by John Wiley & Sons Ltd Additional supporting information may be found in the online version of this article at the publisher’s web site. Keywords: mitochondria; biomarkers; Raman spectroscopy; linear discriminant analysis

Introduction

J. Raman Spectrosc. 2016, 47, 933–939

* Correspondence to: Department of Engineering Sciences and Mathematics, Luleå University of Technology, SE-971 87 Luleå, Sweden. E-mail: [email protected] This is an open access article under the terms of the Creative Commons AttributionNonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. a CBZI, University of Nova Gorica, Glavni Trg 8, SI-5271, Vipava, Slovenia b Department of Engineering Sciences and Mathematics, Luleå University of Technology, SE-971 87, Luleå, Sweden

© 2016 The Authors Journal of Raman Spectroscopy Published by John Wiley & Sons Ltd

933

Among the most conserved aspects of cellular aging is the gradual loss of mitochondrial function.[1,2] By contrast, a high mitochondrial energetic output sustains fast gait speed in nonagenarians.[3] Consequently, understanding mitochondrial function might promote healthy aging. Mitochondria respond to the cell’s energetic needs, environmental conditions and their own homeostasis (self-regulation) by engaging in a dynamic process of fission (division) and fusion. In order to characterize these processes, different yeast mutants have been isolated. Interestingly, the inability to split mitochondria, as reported in yeast dnm1Δ and fis1Δ deletion mutants, leads to a longer and healthier life span.[4] Conversely, fzo1Δ yeast mutants present highly fragmented mitochondria and higher mortality.[5] Fis1 and Fzo1 are integral outer mitochondrial membrane proteins. The dynamin Dnm1, is recruited to mitochondrial constriction sites by Fis1, where it is thought to help pinch the mitochondrial membrane.[5] Instead, the mitofusin Fzo1 is believed to help the joining of mitochondrial tubules. Recently, small molecule inhibitors of Drp1, a homologue of Dnm1, have shown efficacy in reducing the extent of stroke in mice.[6] Finding the mechanism by which mitochondrial dynamics maintains a healthy aged phenotype requires the development of novel detection and analytical tools. In the present study we applied Raman spectroscopy and Linear Discriminant Analysis (LDA) on a concentrated yeast cell suspension as a tool for detecting the mitochondrial signature of healthy longlived and short-lived yeast mutants.[4] Raman spectroscopy is based on photons interacting with molecules. The energy of the photon causes the molecule to vibrate in a unique way and the outgoing photon will consequently have less energy. Because the enerOgy difference of the incoming and outgoing photons is

individual for each molecule, the resulting Raman spectrum gives quantitative and qualitative information on a sample and its state. From its discovery in the late 1920s, Raman spectroscopy has provided biologists with label-free biochemical, morphological and kinetic information. Most Raman-based diagnosis has focused on identifying malignant tissue and human pathogens. Detection of early Alzheimer’s disease using blood plasma has also been proposed.[7] Whole cells represent a complex mixture of overlapping molecular signatures. Nevertheless, Raman spectroscopy has yielded information on septum dynamics, lipid composition, stress response, nuclear division, respiratory activity and cell death in live yeast cells.[8,9] By virtue of their many cis and trans bonds, phospholipids are particularly amenable to spectroscopic characterization. Phosphatidylinositol (PI), phosphatidylserine (PS), phosphatidylethanolamine (PE) and phosphatidylcholine (PC), account for ~80% of mitochondrial membrane lipids,[10] while sterol esters (e.g., ergosterol) are mostly found in lipid droplets. Micro-Raman spectroscopy has also been used to identify the spectral signatures of heme proteins, such as neuroglobin,[11] whose structure is akin to the

N. Erjavec, G. Pinato and K. Ramser cytochromes found in mitochondrial oxidative phosphorylation (OXPHOS) Complexes. During OXPHOS, electrons are transferred from NADH, an electron donor, to Complex I, or NADH:ubiquinone oxidoreductase. From here, they are then passed to Complex III, or ubiquinol cytochrome c oxidoreductase, by ubiquinone. Finally, electrons are transferred by way of cytochrome c, another carrier, to Complex IV, or cytochrome c oxidase, where O2 is reduced to H2O while protons are pumped in the intermembrane space. The redox states of cytochromes b and c,[12] and the vibrational bonds of quinone or NADH have all been detected by Raman spectroscopy. [13 ] The aim of this work was to test the use of Raman spectroscopy and LDA for the detection of chemical differences between longlived dnm1Δ and fis1Δ mutants, and short-lived fzo1Δ and wild type cells, as well as the identification of corresponding biomarkers. Furthermore, to distinguish between respiring and non-respiring mitochondria, we grew cells on either a glycerol or glucose medium, respectively. On glucose, mitochondrial volume is ~3× smaller[13] and energy is generated via glycolysis in the cytosol. Growth on a respiratory carbon source is often used as a discriminatory condition for the study of mitochondrial mutants that are indistinguishable under non-respiring conditions.[14] The spectroscopic assessment of mitochondrial fitness in live cells has the potential of replacing current costly, complex and toxic biochemical assays, reducing the need for large biopsies and long testing times.

Methods Strains and growth conditions Yeast strains used in this study were derivatives of the haploid BY4741 wild type (WT), MATa; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0 (EUROSCARF). Additionally, we used also strains dnm1Δ* (MATa; his3Δ200; leu2Δ0; ura3-52; dnm1Δ::HIS3) and the corresponding wild type WT* (Janet M. Shaw, University of Utah School of Medicine, Salt Lake City, UT, USA). All strains were grown in YP (1% yeast extract, 2% peptone) supplemented with either 2% glycerol (YPG) or 2% D-glucose (YPD) at 29 °C following standard procedures. Competition experiments were started by diluting equal amounts of exponentially growing wild type and mutant yeast cultures and mixing them together in fresh medium. Co-cultures were allowed to grow to OD600 = 1.2, at which point they were diluted 1/10 000 in fresh medium and the procedure was repeated every 24 h. Colony Forming Units (CFUs) were scored by replica plating co-culture aliquots on selective (+G418) and non-selective (YPG or YPD) medium. Competition experiments were performed twice. Microscopy Following manufacturer’s instructions, 106 cells/ml were harvested by centrifugation and resuspended in 1 mL 10 mM HEPES buffer containing 5% glucose. Mitotracker Red (Life Technologies) was added to a final concentration of 100 nM, and cells were incubated in the dark at 25 °C for 20 min. Cells were imaged with a 100× objective on an Axiovert X10 inverted fluorescence microscope (Zeiss). Images were processed with ImageJ software (http://rsb.info.nih.gov/ij/). Raman spectral acquisition

934

A 2-mL aliquot of exponentially growing cells was centrifuged at 22 °C, washed once with PBS and resuspended in PBS + 2% glucose. A drop was placed on a cover glass of Raman inactive glass (Calcium Fluoride UV grade; Crystran Ltd.), which was rinsed and used

wileyonlinelibrary.com/journal/jrs

repeatedly. Spectra were recorded on an inverted microscope (X71; Olympus) with a 40× objective (N.A. = 0.6) coupled by filter optics to a Raman spectrometer (303i; Shamrock). Samples were excited with 1-mW power from a 532-nm Yd:NVO4 laser. The used irradiance of 0.67 kW/cm2 is considered safe, as it does not affect cell cycle progression.[10] About three to six spectra per sample were recorded at different points from the cell suspension, with 50–100 cells per frame. Integration time was 180 s per spectrum. Raman spectral analysis A number of consecutive preprocessing steps were applied to raw spectra: (1) manual removal of cosmic rays with IGOR PRO software, (2) smoothing in R (https://www.r-project.org/) by applying Eilers’ algorithm with λ = 10,[15] (3) background correction by subtracting the buffer spectra, and (4) baseline correction with a polynomial fit over a specified spectral range. Steps 3, 4 and subsequent normalization, averaging, and plotting were performed using the R HyperSpec JSS package (http://hyperspec.r-forge.r-project. org).[16,17] Because the number of variables in an entire spectrum was larger than the number of samples, we reduced spectral dimensionality by Principal Component Analysis (PCA). The scores of the most significant PCs, amounting to a cumulative probability of 95%, were then used as input for LDA.[18] LDA was performed with the R MASS package (http://cran.r-project.org/package=MASS) according to the following steps: (1) calculating the means and covariances relative to each class, starting from multivariate normally distributed spectra, (2) randomizing the data set using the [rnmvnorm] function on a means and pooled covariances matrix, and (3) classifying the samples’ spectra by comparing them to LDA scores of predicted spectra from step 2. PCA was not required for peak-by-peak LDA. Mitochondrial activity measurements Enzymatic activity of respiratory Complexes I, III, and citrate synthase were performed with live yeast cells and crude mitochondrial extracts, as described previously.[19] Briefly, 400 mL yeast culture at OD600 = 1 was harvested, washed and digested with Zymolyase (Zymo Research) followed by 50 strokes in an ice-cold glass Dounce homogenizer (Weather). Mitochondrial pellets were further purified by subsequent rounds of centrifugation,[20] protein concentration was determined by Bradford (Bio-Rad) and pellets were stored at 80 °C until use. Complex I activity followed the decrease in absorbance at 340 nm resulting from the oxidation of NADH, in the absence or presence of rotenone. Complex III activity followed the increase in absorbance at 550 nm resulting from the reduction of cytochrome c, in the absence or presence of antimycin A. Citrate synthase activity followed the decrease in absorbance at 412 nm resulting from the reduction of 5.5′-dithiobis (2-nitrobenzoic acid), in the presence of oxaloacetic acid. Each measurement consisted of three to four repetitions. Citrate synthase measurements were used for normalizing Complex I and III activities.

Results First, respiring dnm1Δ, fis1Δ, fzo1Δ and wild type (WT) yeast cells were stained with a mitochondria specific dye (Mitotracker Red) and observed under the microscope. Under respiratory conditions mitochondria became paramount for energy (ATP) production, grew in volume,[20] and variations in mitochondrial morphology between mutants became more apparent (Fig. 1a). Thereafter, we

© 2016 The Authors Journal of Raman Spectroscopy Published J. Raman Spectrosc. 2016, 47, 933–939 by John Wiley & Sons Ltd

Detecting mitochondrial fitness

Figure 2. Fitness and spectral clustering of long- and short-lived mutants depend on mitochondrial fragmentation. a) Background corrected, smoothed, baseline corrected and normalized spectra for cells grown on glucose. The means of n = 6 (WT), n = 5 (dnm1Δ), n = 5 (fis1Δ), and n = 5 (fzo1Δ) spectra are shown. b) PCA-LDA of the spectra shown above. Actual spectra are represented by crosses, LDA scores of predicted spectra by dots. c) Time resolved fis1Δ to WT ratio of colony forming units (CFUs) from co-cultures grown on glycerol (black symbols) or glucose (grey symbols). Values > 1 indicate a superior competitive fitness of the mutant. One of two repetitions is shown.

recorded the Raman spectra of the various cell types. Considerable differences could be detected between the mutants’ raw Raman signatures in the 400–1850 cm 1 range, as reported in Fig. S1a (Supporting Information). Spectra featured a broad background fluorescence, which was particularly prominent in the 1300– 1850 cm 1 region. Autofluorescence is common in yeast and reflects richness of aromatic compounds, such as tryptophan and nicotinamide. Because of its extent, the 1300–1850 cm 1 ‘fluorescence

region’ could not be included in subsequent analysis; instead, its origin will be discussed in the last paragraph of this section. To assess the significance of the observed spectral differences we limited our analysis to the 400–1300 cm 1 spectral window shown in Fig. 1b. PCA-LDA of single spectra revealed clear clustering of the long-lived dnm1Δ and fis1Δ spectra to one side of the plot, and the short-lived fzo1Δ and WT spectra at a diametrically opposite one (Fig. 1c). In LDA the first discriminant element (LD1) is dominant




935

Figure 1. Raman spectra of yeast mitochondrial mutants cluster into distinct classes, corresponding to their degree of mitochondrial fragmentation. a) Representative whole cell bright field images (upper panel) and fluorescently labeled mitochondria (lower panel) of live yeast cells. Scale bar = 5 μm. b) Background corrected, smoothed, baseline corrected and normalized spectra for cells grown on glycerol. The means of n = 4 (WT), n = 3 (dnm1Δ), n = 6 (fis1Δ), and n = 3 (fzo1Δ) spectra are 1 (1), shown. Specific peaks are indicated and numbered: 510 cm 1 1 1 1 1 750 cm (2), 760 cm (3), 776 cm (4), 875 cm (5), 1258 cm (6). c) PCA-LDA of the spectra shown above. Actual spectra are represented by crosses; LDA scores of predicted spectra by dots. The principal LDA component is indicated as LD1, the second as LD2.

N. Erjavec, G. Pinato and K. Ramser Table 1. Peak assignments and clustering Cellular biomarker Phosphatidylinositol Phosphatidylserine Phosphatidylcholine Phosphatidylethanolamine Sterol esters Glutathione NADH Ubiquinone FMN [Fe―S] Cytochrome b, c, c1 Phenylalanine

1

Non-clustered peaks (cm )

519, 596, 776 524, 733, 787 715, 875, 1092 760

415 1268 429, 702 550

991, 1000 960 1258 460–80, 490–500, 510, 640–60 604, 750

936

over the next one (LD2), and they serve to maximize the betweencluster variance to the within-cluster variance.[18] All spectra (crosses in Fig. 1c) had a probability P = 1 of belonging to the predicted class, defined by the LDA scores of expected spectra (dots in Fig. 1c). Results were independent of strain background, as confirmed by PCA-LDA of the dnm1Δ* mutant and corresponding WT* and presented in Fig. S2 (Supporting Information). In order to verify that clustering was attributable to mitochondrial differences, we repeated the above experiment by growing cells on glucose, as shown in Fig. 2a and Fig. S1b (Supporting Information). Here, PCA-LDA revealed marked spectral overlap among all cell types (Fig. 2b). This was substantiated by a lower degree of separation on the LD1 scale and the spectra being scattered away from the predicted clusters. Respiratory versus fermentative growth was used also to assess the impact of mitochondrial fragmentation on overall cell fitness. In a competition experiment where equal initial amounts of fis1Δ and WT were mixed together, fis1Δ swiftly out-competed the WT on glycerol, but was equally fit on glucose (Fig. 2c). If Raman spectroscopy diagnosis were to depend on identifiable biomarkers, selected peaks would be expected to cluster along the lines of Fig. 1c. To test this hypothesis, we performed LDA on a region of ±5 cm 1 centered around known peaks, some of them indicated in Fig. 1b. As summarized in Table 1, and shown for representative peaks in Fig. 3a–h, clustering was significant for most phospholipids and some OXPHOS components. PI, PC, PE, and PS exhibited a probability P = 1 of belonging to the predicted class (Fig. 3a–d). Conversely, the spectra of sterol esters largely overlapped, and the probability of belonging to the predicted class was generally P < 0.90. Lack of significant clustering for the phenylalanine band was taken to indicate a similar protein content in all cell types. NADH exhibited significant clustering (Fig. 3e) while glutathione, a reducing agent, did not. Clustering was strong for the catalytic [Fe―S] pool of Complexes I and III (Fig. 3f), but not for the flavin mononucleotide (FMN) component of Complex I (Fig. 3g). Similarly, clustering was only partial for the mostly reduced forms of cytochrome b and c1 of Complex III, cytochrome c (Fig. 3h), and ubiquinone (not shown). Resonance Raman scattering for the cytochromes a, a3 of Complex IV could not be achieved with a 532-nm laser. Finally, we performed spectrophotometric assays on crude mitochondrial extracts and living yeast cells to determine the enzymatic activities of Complexes I and III. In line with Raman clustering, activities were high in dnm1Δ and fis1Δ mutants, but failed to differ from WT in fzo1Δ (Fig. 3i–j). Given the elevated NADH:ubiquinone oxidoreductase activity of dnm1Δ and fis1Δ (Fig. 3i), we used Raman spectroscopy to


1

Clustered peaks (cm )

1167 536, 832 1127, 1167 1004

References 30 30 30 30 30 31 31 31 32 33 29,34,35 8

investigate the time-resolved response of YPG grown WT and dnm1Δ cells to the Complex I inhibitor rotenone. Spectra were taken before and at 5-min intervals after addition of a 500-nM sublethal dose of rotenone.[21] The drug caused an immediate spectral change in WT cells, further enhancing the broad fluorescent band at 1300 – 1850 cm 1 and two additional peaks at around 750 and 790 cm 1 (Fig. 4a). This initial spike was followed, within 20 min, by a gradual return towards early values, suggesting that the response had run its course. In comparison, the spectra of dnm1Δ cells, which even before treatment exhibited no fluorescent band, displayed only limited perturbation by rotenone (Fig. 4b), confirming how the drug had its greatest impact on the broad fluorescence peak.

Discussion The scope of this study was to evaluate the feasibility of Raman spectroscopy for the in vivo detection of biomarkers of mitochondrial fitness. Accordingly, we demonstrate that Raman spectra (~50–100 cells/spectrum) can be used to discriminate between live yeast cells with reduced or increased mitochondrial fragmentation. The spectra of less fragmented, long-lived yeasts clustered significantly apart from those of more fragmented, short-lived or WT cells (Fig. 1c). The impact of mitochondrial fragmentation on the Raman signatures of long- and short-lived mutants was confirmed by loss of clustering in glucose grown cells, which depended on fewer mitochondria (Fig. 2b). It should be noted that a metabolic shift from fermentation to respiration is accompanied by, among others, down-regulation of rRNA and up-regulation of stress response pathways. Thus, while the effect on mitochondrial morphology and activity remains dramatic, we could not single out mitochondrial fragmentation as the sole reason for the observed spectral differences between cultures grown on glycerol and glucose. In terms of spectral clustering, it is the unique tubular networks of the long-lived mutants (Fig. 1a) and hence the lack of mitochondrial fragmentation, that differed most from the WT (Fig. 1c). This was confirmed by fis1Δ being fitter than the WT when grown together on glycerol, but not on glucose (Fig. 2c). Consequently, a Raman-based screening would be a valuable tool in identifying environmental factors with an age-retarding effect on mitochondria. In doing so, inherent differences between strains and growth conditions should be taken into account and proper controls should be included. Strain-dependent spectral variation has been reported previously[22,23] and is seen here for dnm1Δ and WT in


Detecting mitochondrial fitness

Figure 3. Peak-specific Raman spectral differences between mitochondrial mutants cluster according to membrane composition and respiratory activity. LDA of representative peaks corresponding to: a) phosphatidylinositol; b) phosphatidylcholine; c) phosphatidylethanolamine; d) phosphatidylserine; e) NADH; f) [Fe―S] clusters; g) FMN; h) cytochromes b, c, c1. Raman shifts are indicated. Color-coding is as in Fig. 1. Enzymatic activity of: i) Complex I; j) Complex III. The means and standard deviations (error bars) of at least three independent experiments are shown.


activity of mitochondria in living yeast.[23,25] The 1602 cm 1 peak was recently ascribed to ergosterol in mitochondria and lipid droplets.[23] Here, no significant clustering was found for the other two ergosterol peaks (Table 1), suggesting that ergosterol was not a discriminant factor in mitochondrial fragmentation. Multiple studies have reported on Dnm1 and Fis1 binding to liposomes containing phospholipids.[10] It is not surprising then that the mutants described in this study, may be affected by vast changes



937

Fig. 1C and Fig. S2b (Supporting Information). Use of complete medium, such as YPG (Fig. 1b) or YPD (Fig. 2a), adds a substantial amount of background fluorescence, particularly in the 1300– 2000 cm 1 region.[22] This may explain why the spectra in this study differ from those obtained on other media, such as SD[9] or lactate.[24] It also means we were unable to evaluate the impact of mitochondrial fragmentation on the Raman spectroscopic signature of life at 1602 cm 1, known to coincide with the respiratory

N. Erjavec, G. Pinato and K. Ramser

Figure 4. Spectral response to Complex I inhibition by rotenone points to NADH autofluorescence. a) WT and b) dnm1Δ cells were incubated in the presence of 500-nM rotenone for 0, 5, 10, 15 or 20 min prior to spectral 1 1 acquisition. Peaks at 750 cm and 790 cm are indicated and a close-up is shown in the inset. Spectra were not normalized.

938

to their membrane make up. While the present work does not reveal variations in abundance of specific lipids, Raman spectral clustering indicates that phospholipids change in relation to mitochondrial fragmentation and life span (Fig. 3a–d). In line with our findings, recent pharmacological studies have shown the importance of the mitochondrial lipidome as a determinant of yeast chronological life span.[26] Accordingly, exogenously added lithocholic acid (LCA), a bile acid, was found to integrate within the inner mitochondrial membrane causing a relative increase in phosphatidic acid, PC, PS, and PI. As a result mitochondria appeared enlarged and fewer in numbers, a phenotype also typical of dnm1Δ and fis1Δ mutants (Fig. 1a). Most strikingly, as in the case of these healthy aged mutants, LCAtreated yeast cells were also characterized by delayed aging. Treatment with LCA was shown to affect fluidity and protein movement through the inner mitochondrial membrane, altering the dynamics of respiration.[26] Our results indicate that the spectral clustering seen for phospholipids applied also to some OXPHOS components (Fig. 3f–h). Thus, we speculate that the healthy aged phenotype of dnm1Δ and fis1Δ is achieved through altered mitochondrial membrane composition, triggering changes in OXPHOS activity. OXPHOS activity was also found to be impaired in psd1Δ yeast mutants, characterized by lower mitochondrial PE levels and lipid mixing.[27] Spectral clustering indicates that [Fe―S] centers of Complexes I and III may constitute reliable biomarkers of healthy aging (Table 1). The fact that not all electron transport components clustered equally may be ascribed to their location in the inner mitochondrial membrane. The FMN moiety of NADH:ubiquinone oxidoreductase faces the matrix, while the [Fe―S] center is largely


membrane bound and may be more responsive to changes in membrane fluidity. This spatial arrangement may explain why [Fe―S] centers, but not FMN, clustered differently in the longlived mutants compared to WT. Because the [Fe―S] center of ubiquinol cytochrome c oxidoreductase is located in the inner membrane space, it may be less affected by membrane changes and have a lesser impact on Complex III activity than the two membrane-bound [Fe―S] centers of Complex I (Fig. 3i–j). Within the aging context, efficiency of the NADH-bound Complex I is particularly important, because of the potential to leak electrons and consequently generate reactive oxygen species. Another possible discriminant of healthy aging is NADH. Not only did NADH cluster significantly in relation to health and life span (Fig. 3e); but it also appeared to give the characteristic fluorescent signal seen in Fig. S1a (Supporting Information). The broad fluorescent band at 1300–1850 cm 1 was likely because of NAD(P)H autofluorescence originating from mitochondria and its less intense cytoplasmic pool.[28] The band was substantially less pronounced in fis1Δ and completely absent from dnm1Δ. The pattern was consistent with a markedly higher NADH:ubiquinone oxidoreductase activity in these cells (Fig. 3i) and a lower NADH/NAD+ ratio. Blocking temporarily the enzyme’s activity with rotenone supported this hypothesis. Uncoupling of respiration caused a momentary excess of NADH, coinciding with a rapid increase in the broad band’s intensity. WT cells displayed an immediate increase of the NADH broad fluorescent band, followed by a gradual return to an unperturbed state (Fig. 4a). Instead, dnm1Δ cells, with their high NADH:ubiquinone oxidoreductase activity, appeared insensitive to the given dose of rotenone (Fig. 4b). Finally, the rapid rise of a peak at around 750 cm 1 in WT cells could be attributed to release of cytochrome c from the mitochondria and thus a sign of apoptosis.[29]

Conclusions In this Raman spectroscopic study we identified phospholipids and OXPHOS Complexes I and III as possible Raman biomarkers of mitochondrial fitness. This was based upon their distinct clustering by LDA in different mitochondrial mutants. Furthermore, the characteristic fluorescent background of NADH could in itself be a vital marker. As a routine prognostic tool, Raman spectroscopy may serve to analyze peripheral blood lymphocytes, skin fibroblasts or skeletal muscle samples for the preventive monitoring of older adults using the biomarkers identified in this study. Acknowledgements We thank Janet M. Shaw, University of Utah School of Medicine, Salt Lake City, UT, USA, for kindly providing yeast strains. We would also like to thank Elsa Fabbretti, University of Nova Gorica and Tanja Dominko, University of Nova Gorica and Worcester Polytechnic Institute, Worcester, MA, USA, for critically discussing the manuscript. The work was supported by grants AHA-MOMENT from the Ministry of Science and Education of the Republic of Slovenia, MINA (Slovenia–Italy 2007–2013 Interregional Grant) and SUNGREEN FP7 RegPot 2011-2015 Grant, the Kempe Foundation and the EU structural fund Norra Norrland Objective II.

References [1] I. R. Lanza, K. S. Nair, Pflugers Arch. 2010, 459, 277.


Detecting mitochondrial fitness [2] M. Scheibye-Knudsen, K. Scheibye-Alsing, C. Canugovi, D. L. Croteau, V. A. Bohr, Aging 2013, 5, 192. [3] P. M. Coen, S. A. Jubrias, G. Distefano, F. Amati, D. C. Mackey, N. W. Glynn, T. M. Manini, S. E. Wohlgemuth, C. Leeuwenburgh, S. R. Cummings, A. B. Newman, L. Ferrucci, F. G. S. Toledo, E. Shankland, K. E. Conley, B. H. Goodpaster, J. Gerontol. A Biol. Sci. Med. Sci. 2013, 68, 447. [4] C. Q. Scheckhuber, N. Erjavec, A. Tinazli, A. Hamann, T. Nyström, H. D. Osiewacz, Nat. Cell Biol. 2007, 9, 99. [5] W. Bleazard, J. M. McCaffery, E. J. King, S. Bale, A. Mozdy, Q. Tieu, J. Nunnari, J. M. Shaw, Nat. Cell Biol. 1999, 1, 298. [6] J. Grohm, S. W. Kim, U. Mamrak, S. Tobaben, A. Cassidy-Stone, J. Nunnari, N. Plesnila, C. Culmsee, Cell Death Differ. 2012, 19, 1446. [7] P. Carmona, M. Molina, M. Calero, F. Bermejo-Pareja, P. Martínez-Martin, A. Toledano, J. Alzheimers Dis. 2013, 34, 911. [8] V. Kaliaperumal, H. O. Hamaguchi, Chem. Record 2012, 12, 567. [9] G. P. Singh, C. M. Creely, G. Volpe, H. Grötsch, D. Petrov, Anal. Chem. 2005, 77, 2564. [10] Q. Zhang, Y. Tamura, M. Roy, Y. Adachi, M. Iijima, H. Sesaki, Cell. Mol. Life Sci. 2014, 71, 3767. [11] K. Ramser, E. Malinina, S. Candefjord, Appl. Spectroscopy 2012, 66, 1454. [12] M. Kakita, V. Kaliaperumal, H. O. Hamaguchi, J. Biophotonics 2012, 5, 20. [13] A. Egner, S. Jakobs, S. W. Hell, Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 3370. [14] N. Erjavec, A. Bayot, M. Gareil, N. Camougrand, T. Nyström, B. Friguet, A-L. Bulteau, Free Rad. Biol. Med. 2013, 56, 9. [15] P. H. C. Eilers, Anal. Chem. 2004, 76, 404. [16] A. Bonifacio, C. Beleites, F. Vittur, E. Marsich, S. Semeraro, S. Paoletti, V. Sergio, Analyst 2010, 135, 3193. [17] C. Beleites, K. Geiger, M. Kirsch, S. B. Sobottka, G. Schackert, R. Salzer, Anal. Bioanal. Chem. 2011, 400, 2801. [18] I. Notingher, G. Jell, P. L. Notingher, I. Bisson, O. Tsigkou, J. M. Polak, M. M. Stevens, L. L. Hench, J. Mol. Struc. 2005, 744–747, 179. [19] A. Barrientos, A. Methods 2002, 26, 307. [20] C. Meisinger, N. Pfanner, K. N. Truscott, Methods Mol. Biol. 2006, 313, 33. [21] A. C. Mesa-Arango, N. Trevijano-Contador, E. Román, R. Sánchez-Fresneda, C. Casas, E. Herrero, J. C. Argüelles, J. Pla, M. Cuenca-Estrella, O. Zaragoza, Antimicrob. Agents Chemother. 2014, 58, 6627.

[22] S. B. Rodriguez, M. A. Thornton, R. J. Thornton, Appl. Environm. Microbiol. 2013, 79, 6264. [23] L. D. Chiu, F. Hullin-Matsuda, T. Kobayashi, H. Torii, H. O. Hamaguchi, J. Biophotonics 2012, 5, 724. [24] L. D. Chiu, M. Ando, H. O. Hamaguchi, J. Raman Spectrosc. 2010, 41, 2. [25] Y. S. Huang, T. Karashima, M. Yamamoto, T. Ogura, H. O. Hamaguchi, J. Raman Spectrosc. 2004, 35, 525. [26] A. Beach, V. R. Richard, A. Leonov, M. Burstein, S. D. Bourque, O. Koupaki, M. Juneau, R. Feldman, T. Iouk, V. I. Titorenko, Aging 2013, 5, 551. [27] E. Y. L. Chan, G. A. McQuibban, J. Biol. Chem. 2012, 287, 40131. [28] M. A. Aon, S. Cortassa, K. M. Lemar, A. J. Hayes, D. Lloyd, FEBS Lett. 2007, 581, 8. [29] M. Okada, N. I. Smith, A. F. Palonpon, H. Endo, S. Kawata, M. Sodeoka, K. Fujita, Proc. Natl. Acad. Sci. U. S. A. 2012, 109, 28. [30] C. Krafft, L. Neudert, T. Simat, R. Salzer, Spectrochim. Acta A Mol. Biomol. Spectrosc. 2005, 61, 1529. [31] A. C. S. Talari, Z. Movasaghi, S. Rehman, I. ur Rehman, Appl. Spectr. Reviews 2006, 50, 46. [32] M. Hikita, K. Shinzawa-Itoh, M. Moriyama, T. Ogura, K. Kihira, S. Yoshikawa, Biochemistry 2013, 52, 98. [33] M. Gomez, R. V. Peréz-Gallardo, L. A. Sánchez, A. L. Díaz-Pérez, C. Cortés-Rojo, V. Meza Carmen, A. Saavedra-Molina, J. Lara-Romero, S. Jiménez-Sandoval, F. Rodríguez, J. S. Rodríguez-Zavala, J. Campos-Garcia, PLoS One 2014, 9e111585. [34] N. A. Brazhe, M. Treiman, B. Faricelli, J. H. Vestergaard, O. Sosnovtseva, PLoS One 2013, 8e70488. [35] N. A. Brazhe, M. Treiman, A. R. Brazhe, N. L. Find, G. V. Maksimov, O. V. Sosnovtseva, PLoS One 2012, 7e41990.

Supporting information Additional supporting information may be found in the online version of this article at the publisher’s web site.

939


