Label-free assessment of replicative senescence in mesenchymal stem cells by Raman microspectroscopy

: Here, Raman microspectroscopy was employed to assess replicative senescence of mesenchymal stem cells (MSC). A regular spectral change related to the cell senescence was found in the ratio of two peaks at 1157 cm − 1 and 1174 cm − 1 , which are assigned to C-C, C-N stretching vibrations in proteins and C-H bending vibrations in tyrosine and phenylalanine, respectively. With the cell aging, the ratio I 1157 / I 1174 exhibited a monotonic decline and showed small standard deviations, so that it can statistically distinguish between cells having slight changes in terms of aging. We propose that I 1157 / I 1174 can act as a characteristic spectral signature for label-free assessment of MSC senescence.


Introduction
Mesenchymal stem cells (MSCs), which are present in various tissues such as bone marrow, adipose, amniotic membrane, etc., have been proposed to be a promising candidate for regenerative medicine [1] because of their multi-differentiation potential, hematopoieticsupporting activity, immunomodulatory potency and self-renewal. Umbilical cord is an important source of MSCs [2][3][4][5]. Compared with those derived from bone marrow, MSCs from human umbilical cord (hUC-MSCs) can effectively avoid the reduction in cell number, proliferation and differentiation capacity with age or age-related disease. Therefore, hUC-MSCs are considered to be an ideal source for cell-based therapy.
Generally, cell subculture in vitro is essential to provide a sufficient number of MSCs for a successful clinical application [6]. However, this process is inescapably accompanied by replicative senescence, which has been found to inhibit cell proliferation [6], impair cell differentiation [7], and further abrogate cells' therapeutic potential [8]. Thus, to assess the cell senescence is of great necessity for their therapeutic use.
Empirically, one can roughly estimate cell senescence according to the times of cell passage. However, due to individual differences in cells, there is no strict correspondence between the times of cell passage and the level of cell aging, and thus easy to cause a large error. In addition, cell aging is accompanied with morphological changes such as enlarged cell size and flattened morphology [9][10][11]. Nevertheless, morphological methods are susceptible to interference of individual judgment, and also difficult to quantitatively determine the cell senescence. Furthermore, molecular changes associated with cell senescence have been widely studied, and several meaningful biomolecular markers were revealed. For instance, β-galactosidase has been widely accepted as an effective indicator for cell aging detection [9][10][11][12]. The expression of the proteins p16, p21, p53, etc. were also found closely related to MSC senescence [12][13][14]. Therefore, corresponding to the molecular markers, biochemical methods including SA-β-gal Staining, Real Time PCR, Western Blotting, etc. can offer relatively accurate results and make quantitative analysis. However, most of these methods need labeling or complicated operations, which may not be conducive to detection of a large number of cell samples. Hence, development of a label-free optical method for cell senescence assessment is necessary for MSCs' quality control and their applications.
In this regard, Raman spectroscopy may be a promising candidate to tackle the task. This technology is based on inelastic light scattering, in which the frequency of the scattered light is changed as compared to that of the incident light, and their frequency differences are dependent on the molecular vibrations. Therefore, Raman spectroscopy is able to illustrate the intrinsic molecular structures [15] and make quantitative analysis of molecular changes in biological samples [16]. Owing to the excellent features of label-free and "molecular fingerprint", this technology has been increasingly employed in stem cell studies, such as discrimination of undifferentiated and differentiated embryonic stem cells (ESCs) [17][18][19], distinguishing between MSCs and fibroblasts [20], identification of abnormal stem cells [21], in situ monitoring of the MSCs' osteogenic differentiation [22,23] and the cardiac differentiation of ESCs [24], determination of the cell cycle phase in ESCs [25], as well as rapid cell sorting of cardiomyocytes derived from ESCs [26]. In our previous works, the Raman spectral markers specific to viability of hUC-MSCs and their correlations with reactive oxygen species (ROS) were also revealed [27,28]. These encouraging findings showed a tremendous application potential of Raman spectroscopy in living cell analysis, and laid the foundation for its further use in stem cell researches.
In this study, we focused on the Raman spectral variations of hUC-MSCs during their aging process, aimed at looking for the specific Raman spectral features that can be linked to the cell replicative senescence. The findings may be helpful to provide new evidences in terms of molecular structures for the cell aging studies, and also contribute to development of a label-free method for assessment of MSC senescence.

Isolation and culture of hUC-MSCs
We isolated hUC-MSCs from human umbilical cord tissues by the method previously described [4]. hUC-MSCs were cultured in monolayer at 37 °C and 5% CO2 in DMEM with 10% FCS. Cells were harvested after reaching 90% confluence using trypsin and passaged (subplated) at a 1:3 ratio until passage (P) 30. We used cells in passage 5, 10, 20, 30 respectively for experiments to obtain samples at different aging levels.

Raman microspectroscopy
A Raman microspectrometer (LRS-5, GANGDONG, China) was used to record the Raman scattering from the center of single hUC-MSCs by means of the microscopic imaging. Briefly, excitation light at 532 nm was provided by a Nd:YAG laser with the output power of 8 mW. A 50 × objective was used to focus the excitation light and to collect the Raman signal. For each spectral scan, Raman spectrum was recorded in the range of 600-1800 cm −1 , and the signal was integrated for 50 s. For each cell sample, 6 ~7 cells were selected randomly for Raman spectroscopy measurements.

Spectral processing and statistical analysis
The background signal was subtracted from each spectrum, and a multipoint baseline was corrected in the whole spectral region. Spectra were then normalized to the peak intensity at 1449 cm −1 that was widely accepted as a normalization standard of Raman spectra from cells [18,29]. The statistical analyses were performed by the Student's t-test.

Results
In order to find the spectral differences related to MSC senescence, we first compared the Raman spectra of cells in passage 5 (P5) with those in P30. As is shown in Fig. 1, obvious spectral differences can be found in the regions of 905 -1016, 1087 -1190, and 1510 -1530 cm −1 . These spectral regions include Raman peaks at 934, 980, 1003, 1097, 1127, 1157, 1174, and 1521 cm −1 , which are mainly corresponding to vibrations in proteins, lipids and DNA [18,27,29,30]. The assignments of the Raman peaks are listed in Table 1.  The spectral differences showed above were observed from average spectra between young and aging cell-groups. However, individual differences among cells in the same group were not taken into account. If individual differences are too large, the corresponding differences from average spectra will be meaningless. Thus, each Raman peak listed above needs to be further validated by statistical analysis. Here, Student's t-test is employed and its result is described by the p value. As a rule of thumb, p<0.05 represents the difference is of statistical significance. The peak intensity comparisons between young (P5) and aging (P30) hUC-MSCs are illustrated in Fig. 2. It can be seen that the average intensities of Raman peaks at 1003, 1127, 1174 cm −1 did not show obvious differences between the two cell-groups and thus cannot be used as spectral markers. As for the peaks at 934, 980, 1097 and 1521 cm −1 , although their average intensities exhibited visible differences, these differences were not statistically significant (p>0.05) because of the relative large standard deviations. In contrast, the Raman peak at 1157 cm −1 , corresponding to C-C, C-N stretching vibrations in proteins, showed significant difference in peak intensity as indicated by p<0.05. In comparison of the case of young cells, the peak intensity of 1157 cm −1 reduced by about 60% in the spectra of aging cells, indicating this peak is sensitive to replicative senescence of hUC-MSCs. However, the results above are not enough to demonstrate the Raman peak at 1157 cm −1 can be used as a spectral marker for the cell aging assessment. It is noted that a spectral marker should not only show a significant difference between young and aging cells, but also exhibit a monotonic change with cell senescence. Thus, we further analyzed Raman spectra from hUC-MSCs in P5, P10, P20 and P30, corresponding to increasingly serious aging states. As is shown in Fig. 3, average spectra of the cells in P5, P10, P20, P30 are visually compared and the Raman peak at 1157 cm −1 is highlighted by a gray bar. Figure 4(A) shows the change of peak intensity at 1157 cm −1 with the cell senescence. It can be seen that the average peak intensity underwent a monotonic decline as the cell passage increased. However, this decline showed significant nonlinearity i.e. sharp decline could be observed from P5 to P10, while only a slight decrease occurred from P10 to P30. This is not conducive to the characterization of the cell aging. More disadvantageously, for each cell group, the peak intensity exhibited so large standard deviation that the p values obtained from Student's t-test between cells in P5 and P10, P10 and P20, P20 and P30 were all larger than 0.05. Statistically, it means the peak intensity at 1157 cm −1 does not have the ability to detect slight changes in the cell aging state. Values p 1 , p 2 , p 3 were obtained by the Student's t-test analysis between cells in P5 and P10, P10 and P20, P20 and P30 respectively.
Since no single Raman peak can be an effective indicator for the cell aging assessment, multi-peak complex should be considered. In this respect, we try to find a composite indicator by calculation of two or more peak intensities. As is shown in Fig. 3, a monotonic change in the ratio of peak intensities at 1157 cm −1 and 1174 cm −1 can be clearly observed and highlighted by the gray bar. Further, the change of ratio I 1157 / I 1174 with the cell aging and the results of Student's t-test are shown in Fig. 4(B). Compared with the case of I 1157 (Fig. 4(A)), the ratio I 1157 / I 1174 showed a more linear decline with the cell passage increase, especially from P10 to P30. More importantly, the standard deviations of I 1157 / I 1174 in the cell groups P5 and P10 were much smaller, indicating the ratio is more stable, and less affected by individual differences of the cells. Statistically, the p values by Student's t-test of I 1157 / I 1174 between cells in P5 and P10, P10 and P20, P20 and P30 were all less than 0.05, suggesting I 1157 / I 1174 is potential to detect slight changes in the cell aging state. Besides, It is necessary to mention that this result can be verified by other cell lines of hUC-MSCs derived from different donors. In those cases, it can also be observed that I 1157 / I 1174 of young cells was significantly higher than that of aging cells (data not shown). Based on these findings, we suggest that I 1157 / I 1174 can be an effective spectral marker for assessment of MSC senescence.

Discussion
Similar to any normal somatic cell, MSCs undergo replicative senescence and have a limited lifespan in vitro [6,9]. It has been reported that replicative senescence of MSCs is a continuous process, during which numerous changes in phenotype, differentiation potential, global gene expression patterns, and miRNA profiles occur [31]. These far-reaching alterations will inevitably affect the cell therapy and other MSC applications. Thus, cell senescence is an important factor that needs to be considered and evaluated in MSC preparation, especially through a label-free manner. In this work, we found a Raman spectral feature (I 1157 / I 1174 ) that showed high sensitivity and effectiveness for MSC aging assessment, indicating Raman spectroscopy can be an ideal choice. Moreover, owing to the "molecular fingerprint" feature, Raman spectroscopy also showed a great potential in monitoring MSC viability [27] and differentiation [23]. These encouraging findings make it hopeful to further develop a label-free method for monitoring multiple physiological parameters of MSCs.
In addition, the spectral alterations may provide new evidences for understanding of replicative senescence in MSCs. The results showed that significant spectral difference between young and aging cells can be found at 1157 cm −1 (Fig. 2), which is assigned to C-C, C-N stretching vibration in peptide bonds and sensitive to the conformational changes of proteins [30]. With the cell aging, the peak intensity of 1157 cm −1 exhibited a significant decline (Fig. 4), while no alterations in peak position and full width at half maximum (FWHM) were observed, indicating a damage of the peptide bonds. The changes in protein structures were also indicated by the Raman peak at 934 cm −1 that is assigned to C-C backbone stretching vibrations in proteins. As the cell passage increased, an obvious expansion in the peak FWHM of 934 cm −1 can be observed (Fig. 3), indicating a change in the protein backbones. These results suggest that protein cleavage may occur in the cell aging process.
Besides, the change of peak at 1174 cm −1 should also be discussed. Although there is no statistically significant difference in this peak between young and aging cells can be found (Fig. 2), it is noted that the standard deviation of the peak intensity ratio I 1157 / I 1174 was much smaller than that of single peak at 1157 cm −1 , and the ratio showed a much better performance in distinguishing cells at different aging states (Fig. 4). In other words, the ratio I 1157 / I 1174 largely reduced the fluctuations of I 1157 caused by individual differences and showed a greater ability for assessing the cell senescence. It indicates that for a single cell, the rise of I 1174 occures concomitantly with the decrease of I 1157 during the cell aging process. The Raman peak at 1174 cm −1 is corresponding to C-H bending vibrations in tyrosine and phenylalanine [18], which are cellular endogenous fluorescent substances. Thus, the rise of I 1174 with the cell senescence was further confirmed by autofluorescence measurement. As is shown in Fig. 5, the fluorescence of aromatic amino acids (emission peak at 336nm, excitation wavelength at 280nm) in aging cells (P20) was higher than that of young cells (P5), and their statistically significant difference was observed (p<0.05). Taken together, we infer that the aging of hUC-MSCs is accompanied by cleavage of proteins and polypeptides, and thus free amino acid residues will increase in the senescent cells. In this case, more amino acid residues are exposed to irradiation, causing an enhancement of their Raman scattering and fluorescence intensity. These molecular changes may be closely related to oxidation reactions and oxidation end products, which have been proven to play an important role in MSC senescence [32].
Finally, it is necessary to mention that the sample size is relatively small, due to the limited cell lines and the long-term cell passage. The small sample size may introduce somewhat influences on the results caused by systematic drifts and baseline corrections, etc., which are almost unavoidable in such weak signal measurements and potentially lead to minor deviations in the results. Future work based on an expanded sample size is needed to further explore quantitative correspondence between the Raman spectral feature and MSC senescence.

Conclusions
In this study, we demonstrated that the C-C, C-N stretching vibrations in proteins at 1157 cm −1 and C-H bending vibrations in tyrosine and phenylalanine at 1174 cm −1 can be associated with the replicative senescence of hUC-MSCs, and their intensity ratio I 1157 / I 1174 can act as a characteristic spectral signature for label-free assessment of the cell senescence. The decrease of I 1157 / I 1174 with the cell aging could be related to the cleavage of proteins and increase of free amino acid residues in the senescent cells.