Autofluorescence is a Reliable in vitro Marker of Cellular Senescence in Human Mesenchymal Stromal Cells

Mesenchymal stromal cells (MSC) are used in cell therapies, however cellular senescence increases heterogeneity of cell populations and leads to uncertainty in therapies’ outcomes. The determination of cellular senescence is time consuming and logistically intensive. Here, we propose the use of endogenous autofluorescence as real-time quantification of cellular senescence in human MSC, based on label-free flow cytometry analysis. We correlated cell autofluorescence to senescence using senescence-associated beta-galactosidase assay (SA-β-Gal) with chromogenic (X-GAL) and fluorescent (C12FDG) substrates, gene expression of senescence markers (such as p16INK4A, p18INK4C, CCND2 and CDCA7) and telomere length. Autofluorescence was further correlated to MSC differentiation assays (adipogenesis, chondrogenesis and osteogenesis), MSC stemness markers (CD90/CD106) and cytokine secretion (IL-6 and MCP-1). Increased cell autofluorescence significantly correlated with increased SA-β-Gal signal (both X-GAL and C12FDG substrates), cell volume and cell granularity, IL-6/MCP-1 secretion and with increased p16INK4A and CCND2 gene expression. Increased cell autofluorescence was negatively associated with the expression of the CD90/CD106 markers, osteogenic and chondrogenic differentiation potentials and p18INK4C and CDCA7 gene expression. Cell autofluorescence correlated neither with telomere length nor with adipogenic differentiation potential. We conclude that autofluorescence can be used as fast and non-invasive senescence assay for comparing MSC populations under controlled culture conditions.

cells is associated to decreased tissue regeneration capacity and life expectancy, and their elimination in a mouse model resulted in increased lifespan 16 . This identifies cellular senescence as an ideal target for the development of new anti-ageing therapies. Nevertheless, interventions and detection of senescent cells, both in vitro and in vivo, need further optimizations because of the complicated nature of the ageing process.
Senescence is a multifactorial process and the heterogeneity of cell populations can hinder the identification of specific markers. The most commonly used laboratory assay for the detection of cellular senescence is senescence-associated beta-galactosidase (SA-β-Gal) 17 , however the method has some limitations. Confluency, serum starvation and operator bias influence the assay outcomes with false positive results 18 . False negatives are also an issue -where some cells enter senescence without detectable SA-β-Gal activity 19 . Several other markers -such as oxidative status 20 -can be used to identify cellular senescence, nonetheless their determination is time consuming and technically difficult to assess in living tissues. Therefore, development of readily available biomarker and the development of a sensitive and easy-to-perform assay to identify cellular senescence would provide major benefits in a rapidly evolving and challenging field such as regenerative medicine.
The first observations of autofluorescence were reported over 100 years ago 21 , as a spontaneously occurring phenomenon due to the accumulation within the cytoplasm of endogenous molecules with fluorophore-like properties. In senescent cells, under the pressure of oxidative stress, the level of autofluorescence increases because the aggregation within the cytoplasm of lipofuscin and lipofuscin-like compounds 22 . Analysis of cellular autofluorescence can be quickly and extensively used as it relies on common flow cytometry instruments (or fluorescent microscopy) which detect excitation of endogenous fluorophores produced by cell constituents, such as oxidative chain respiratory proteins (nicotinamide adenine dinucleotide (NADH), flavin adenine dinucleotide (FAD) and cytochrome C), structural proteins (collagen and elastin), vitamins (riboflavin and thiamine) and undegredable waste material (lipofuscin) 21 . Lipofuscin has been referred to as the "age pigment" 23 , as it accumulates within the cytoplasm with age and correlates with the rate of ageing 24 . Furthermore, riboflavin has already been recognized as a biomarker associated selectively with epithelial cancer stem cells 25 , while fluorescence has been proposed as a non-invasive indicator of embryonic stem cell differentiation state, to distinguishing pluripotent from more mature cells 26 .
Based on the above facts, we speculated that cellular autofluorescence would be an ideal candidate to test for the purpose of tracking senescence in MSC. The co-localization of lipofuscin and SA-β-Gal activity in senescent cells both in vitro and in vivo has been demonstrated in archival tissues, supporting the idea of using lipofuscin as biomarker for cellular senescence 27 , however no study has been conducted to elucidate whether the autofluorescence of MSC could be linked to measures of cellular senescence. Cellular senescence has been successfully assessed not only by SA-β-Gal assay with chromogenic (X-GAL) 17 and fluorescent (C 12 FDG) 28,29 substrates, but also by cell size 30 and granularity 31 , secretion of senescence-associated cytokines (IL-6 and MCP-1) 32 , gene expression of cell cycle regulators associated to cell senescence (p16 INK4A , p18 INK4C , p21 CIP1 , E2F1, ANKRD1, CCND2, CDC2 and CDCA7) [33][34][35][36] and telomere length 37 . Variations in MSC stemness linked to cell senescence are monitored by surface markers (CD90 and CD106) 20,38 and differentiation potential by adipogenic, chondrogenic and osteogenic assays 39 . In the present study, we tested the suitability of an autofluorescence profile of bone marrow-derived MSC measured by flow cytometry, as a tool for a rapid and non-invasive prediction of MSC senescence in correlation with the above mentioned markers for senescence, stemness and differentiation. We also included in the study three different culture conditions and extended our analysis to adipose-derived MSC and peripheral blood lymphocytes.

Results
Correlation of cellular senescence to autofluorescence in mesenchymal stromal cells (MSC). In order to characterize cellular senescence, bone marrow isolated MSC were initially categorized by their senescence-associated beta-galactosidase (SA-β-Gal) activity, evaluated with chromogenic (X-GAL, Fig. 1a) and fluorescent substrates (C 12 FDG, Fig. 1b). The proportion of X-GAL positive cells, as a percent of the total population, significantly increased with cellular autofluorescence (τ b = 0.672, p < 0.001), as well as the fluorescence intensity of MSC population stained with C 12 FDG (τ b = 0.703, p < 0.001). Interestingly, we also observed a high degree of correlation between cellular granularity (Fig. 1c) -represented by side scatter values -and cellular autofluorescence of MSC (τ b = 0.839, p < 0.001). Representative images of samples characterized by low, middle and high degree of senescence stained by X-GAL (Fig. 1d) and C 12 FDG (Fig. 1e) are shown.
Senescent MSC show well defined morphological alterations, such as enlarged and flattened aspect 40 (Fig. 2a), and for this reason cellular autofluorescence was also correlated to their size, as a measure of cell volume (Fig. 2b) and forward scatter values (Fig. 2c). A strong and significant correlation was found between autofluorescence and both cell volume (τ b = 0.698, p < 0.001) and forward scatter values (τ b = 0.776, p < 0.001).
To further explore the dynamics of MSC during the process of cellular senescence, we evaluated other senescence-associated markers such as changes in gene expression, cytokine secretion and telomere length. At the gene expression level, we observed a low and non-significant correlation between the expression of p16 INK4A ( Alternative variables which we considered that might be involved in replicative cellular senescence were increasing donors' age ( Fig. 4c) and telomere shortening (Fig. 4d). Including only low passage MSC populations (from P1 to P4), we observed a low degree and non-significant correlation between donors' age and cellular    Correlation of cellular senescence and autofluorescence over increasing passages in MSC cultures. Changes in cellular senescence were assessed by comparing early (P1 to P4) and late (P6 to P15) passages of MSC (Fig. 6). Compared to early passages, late passages had significantly increased cellular autofluorescence (p = 0.012), fluorescent (C 12 FDG, p = 0.011) and chromogenic (X-GAL, p = 0.012) SA-β-Gal activity, and cellular granularity (SSC, p = 0.027). Cell size, as a measure of cell volume (p = 0.011) and forward scatter values (p = 0.012), were also significantly increasing with passages. While the proportion of cells positive to CD90/ CD106 staining decreased significantly (p = 0.028), telomere length was decreasing non-significantly with time in culture. MSC gene expression of CDCA7, CDC2, p18 INK4C and E2F1 (for all four, p = 0.012) was significantly decreasing at late passages compared to early ones. On the other hand, the expression of p16 INK4A , CCND2, ANKRD1 and p21 CIP1 was increasing with cell passaging (data not shown), however only changes in p16 INK4A expression were significant (p = 0.017). The adipogenic potential of early and late passage MSC was equivalent, while chondrogenic (p = 0.035) and osteogenic (p = 0.012) potentials were significantly reduced at late passages.
Although not significantly, the secretion of IL-6 and MCP-1 was increasing at later passages (data not shown). Protein extracts from early (low autofluorescence) and late (high autofluorescence) passages MSC were used to assess changes in lipofuscin expression with GGH, FAM96B and HDGFL1 antibodies 41 by western blot (Fig. 7a). Quantification of bands ( Fig. 7b) showed that -in comparison to low autofluorescent MSC -high autofluorescent MSC (n = 3) had higher expression of GGH (1.3-fold), of FAM96B (1.9-fold) and of HDGFL1 (1.3-fold).
Correlation of SA-β-Gal activity and autofluorescence in different culture conditions. We evaluated the influence of glucose concentration and oxygen levels on the correlation between cellular autofluorescence and SA-β-Gal. MSC (n = 6) were grown and sampled after 3, 5 and 7 days in high glucose DMEM/Ham's F12 (4.5 g/L glucose, Amimed) and normoxia (~21% O 2 ), growing medium DMEM/Ham's F12 (3.1 g/L glucose) and hypoxia (5% O 2 ), and growing medium DMEM/Ham's F12 (3.1 g/L glucose) and normoxia (control). In the control culture the proportion of C 12 FDG positive cells strongly correlated with cellular autofluorescence (τ b = 0.830, p < 0.001), and to a lower extent under hypoxic (τ b = 0.673, p < 0.001) and high glucose(τ b = 0.634, p < 0.001) conditions (Fig. 8). Interestingly, while SA-β-Gal intensity was similar in all groups, MSC autofluorescence of cells exposed to high glucose levels was twofold higher compared to control (p < 0.001), but unaltered in hypoxia.

Correlation of SA-β-Gal activity and autofluorescence in other cell types.
To evaluate if the correlation between SA-β-Gal activity and autofluorescence was cell-type specific, we performed the same analysis in two other cell types, namely peripheral blood lymphocytes (Fig. 9a) and adipose-derived MSC (Fig. 9d). In lymphocytes, after the exclusion of dead cells (Fig. 9b), C 12 FDG fluorescence intensity had low and non-significant correlation to autofluorescence ( Fig. 9c; On the contrary, in living adipose-derived MSC (Fig. 9e), C 12 FDG fluorescence intensity was strongly and significantly correlating to autofluorescence ( Fig. 9f;

Discussion
Ageing is an irreversible process in multicellular organism and only recently it has been mathematically proven to be unavoidable, by demonstrating the fine balance regulating accumulation of senescent cells in tissues and prevention of cancer cells proliferation 14 . However, it is not trivial to find the link between chronological age and biological senescence, and for individual organisms it requires empirical measurements. In this study, we showed the successful use of endogenous autofluorescence as a non-destructive tool to determine the in vitro age of mesenchymal stromal cells (MSC).
When cells become senescent in vitro they significantly change their phenotype. These changes also affect MSC deeply at a functional level. For instance, late passage MSC were clinically shown to be less effective in ameliorating graft-versus-host disease than early passage cells 7,42 . Therefore, a need exists to develop reliable and efficient tools to monitor the progression of senescence in tissues and cells -especially in the context of tissue regeneration -for the proper assessment of efficacy, patient inclusion criteria and regulatory/insurance issues.
Based on the literature, we speculated the involvement of lipofuscin in the autofluorescence observed by us, as it is mostly undegredable and accumulates only in senescent or slowly dividing cells, while proliferating cells dilute lipofuscin content with progressive divisions 43 . Indeed, Western blot analysis showed a significant higher accumulation of lipofuscin-related proteins -FAM96B and GGH -in highly autofluorescent MSC, compared to low ones. The different degree of antibody affinity to lipofuscin -which is a well-known molecule, but still elusive in its structure -might be the cause of the non-significant results with HDGFL1 antibody. Lipofuscin autofluorescence in MSC is localized within lysosomes 44 and the increase in lysosomal mass occurring with senescence 45 , is partially reflected by increased expression of senescence-associated beta-galactosidase (SA-β-Gal) 17 . Considering that our results show increased autofluorescence and granularity correlating to MSC senescence, we assumed a direct link between autofluorescence and elevated lysosome content or activity, as a surrogate marker for SA-β-Gal activity 18 , which in turn can be used to assess, with a label-free method, cellular senescence in MSC.
Considering the strong reported links between lipofuscin autofluorescence and individual cell dysfunction, we presumed that 'the missing link' that will mark an advance in the field of high throughput detection of MSC senescence is the use of a sensitive, precise and easy to automate tool, such as flow cytometry. To validate such an approach, we demonstrated the strong association between autofluorescence and commonly used markers of senescent cells in vitro. Most importantly, we observed that increased expression of SA-β-Gal 17 -measured both with chromogenic and fluorescent substrates -was strongly correlating with increased autofluorescence. Similarly, we also found a strong positive correlation with cell size of senescent MSC (as they turn into flat and hypertrophic cells), the expression of genes commonly expressed by senescent cells 33    to autofluorescence, confirming previous findings of declining differentiation potentials after prolonged time in culture 30 . The expression of two stemness surface markers 20,38 -CD106 (VCAM-1) and CD90 -here shown to be downregulated with cell senescence -showed only a weak correlation with autofluorescence. Although telomere erosion is a well-established marker of cell senescence, inter-donor variability renders it an unreliable measure of MSC senescence 20,30,46 . Rather, "cell culture stress" such as oxidative stress (hyperphysiological levels of O 2 ), variability between serum and media batches, and suboptimal culture factors has a more remarkable effect on cell division arrest 47 , where senescence results from. With the progression of senescence, MSC accumulate cell debris and excess of actin fibers promoting the formation of a granular cytoplasm 31 -demonstrated by significantly higher scatter values -and thus increased autofluorescence. The same effect of cell debris accumulation and increased autofluorescence were evident not only in bone marrow derived MSC but also in adipose-derived MSC (ADSC). The absence of correlation between lymphocyte autofluorescence and SA-β-Gal activity only shows that such phenomena are cell type specific. Possibly, lymphocytes are too short-lived to accumulate the intracellular material which characterized cultured MSC and ADSC, or simply they have lower overall cytoplasm volume and a different ageing cell metabolism.
Despite the obvious advantages, there are some limitations in the use of autofluorescence. For example, the low correlation between some senescent markers -such as gene expression of p16 INK4A and p21 CIP1 -and autofluorescence might be result of high inter-donor variability. In this study we used the most common laboratory practices to culture MSC, but we do not know if factors such as confluence and serum starvation, reported as limiting conditions for the SA-β-Gal assay, are likely to affect the autofluorescence signal. For instance, we observed that cell death affected autofluorescence outcome of MSC populations. Dead/dying cells had a significant lower autofluorescent levels compared to living cells and for this reason we included only living cells in our cell fluorimeter analysis. Also, similar autofluorescence changes observed with cellular ageing may be compromised from other perturbations, such as different media components. We observed that neither higher level of glucose in culture (4.5 g/L) nor hypoxia (5% O 2 ) influenced the expression of SA-β-Gal (C 12 FDG fluorescence intensity), but rather affected cellular autofluorescence. However, the correlation between cellular autofluorescence and C 12 FDG fluorescence intensity was consistent among samples grown in the same culture conditions. We concluded that autofluorescence, as an indicator of MSC senescent level, measured in different culture settings cannot be compared directly. Furthermore, given the variety of assays we used in this project to analyze autofluorescence, we had to make a choice and omitted other existing senescent markers, such as DNA damage response (DDR) 48 -measuring the genomic damage occurring at non-telomeric sites -and cell autophagy 49 .
In conclusion, we demonstrated that autofluorescence is a prospective indicator of MSC senescent level in a controlled culture setting. Considering the declining functioning of MSC with age and prolonged in vitro culture, the request for reliable, fast and label-free cell characterization is fundamental for the assessment of impairment of tissue homeostasis and tissue repair, and ultimately the difference between successful regeneration or organ failures 50 . Despite the inter-donor heterogeneity of MSC populations, autofluorescence can be successfully used to indicate the overall potential of cell populations. The assessment of MSC quality in clinical applications requires documentation at every step between MSC extraction, expansion and administration. We believe that this label-free detection of cell senescence can contribute not only to a better understanding of the properties of MSC, but has the potential to become a standard MSC senescence marker. Building on our previous work on assessing MSC ageing and differentiation potential 20,30,51 , we conclude that the measurement of autofluorescence is an economic assay which can be used alone, or complementary to other tests, in order to improve research and the outcomes of cell-based therapies. In perspective, further development of autofluorescence as a validated senescence marker would allow improved sensitivity and specificity measurements in cells -as done with validated diagnostic biomarkers -but in order to so there is a need for standardized cell senescence reference biomarkers.

Materials and Methods
All methods were performed in accordance with the relevant guidelines and regulations of this journal.
Frozen MSC (n = 24) at different passage number (ranging from P1 to P15; for 8 donors, both early and late passages were included) were thaw and cultured for two days (SA-β-Gal assay), three days (cytokine measurements) and four days (phenotypic characterization, RNA and DNA isolation) in growing medium. Cell volume was determined by Scepter cell counter (Millipore, Thermo Fisher Scientific, Zug, Switzerland).
In addition to the previously described growing medium, MSC isolated from six donors were also cultured in diverse conditions to compare their influence on cell autofluorescence. MSC were grown either in high glucose DMEM/Ham's F12 (4.5 g/L glucose, Amimed) and normoxia (~21% O 2 ), or in growing medium DMEM/Ham's F12 (3.1 g/L glucose) and hypoxia (5% O 2 ). For each donor (n = 6), cells in the respective cultures were sampled after 3, 5 and 7 days and results were compared to growing medium in normoxia (control).
Adipose marrow-derived MSC (ADSC). Adipose tissue, in the form of liposuction or excision samples, was obtained from 6 healthy donors following informed consent and according to a protocol approved by the local ethical committee (EKBB, Ref. 78/07). All donors were females between 28 and 75 years of age. Finely minced excision samples or liposuction samples were digested with 0.075% collagenase type II (355 U/mg, Worthington, Lakewood, NJ, USA) for 60-90 min at 37 °C, as previously described 52 . After centrifugation at 190 g for 10 min, the lipid-rich layer was discarded and the cellular pellet was washed once with PBS. Red blood cells were lysed by incubation for 2 min in a solution of 0.15 M ammonium chloride, 1 mM potassium hydrogen carbonate (both Merck, Darmstadt, Germany) and 0.1 mM EDTA (Fluka Analytical, Sigma). The resulting SVF cells were then resuspended in α-MEM supplemented with 10% FBS, 1% HEPES, 1% Sodium pyruvate and 1X of Penicillin-Streptomycin-Glutamine solution (all three from Gibco) at 37 °C in a humid atmosphere containing 5% CO 2 . SVF cells were seeded at a density of 2 × 10 3 cells/cm 2 and cultured in medium supplemented with 5 ng/ mL bFGF, serially replated in new dishes when reaching subconfluence, and then frozen at −150 °C.
In the fluorogenic method, cells were pre-treated for 1 hour with 100 nM bafilomycin A1 (Sigma), which inhibited lysosomal acidification, and then incubated for 1 hour with 20 µM 5-dodecanoylaminofluorescein di-β-D-galactopyranoside (C 12 FDG, Sigma) a fluorogenic substrate for β-galactosidase. Cells were then detached with prewarmed 0.05% trypsin-EDTA and resuspended in PBS containing the viability stain TO-PRO-3 (Molecular probes). After 15 minutes, the measurement of cell fluorescence without (autofluorescence) and with C 12 FDG was carried out with flow cytometer (15′000 events). Data were acquired using the excitation laser at 488 nm and detection optic at 525/50 nm for both autofluorescence and C 12 FDG. A 638 nm laser and 670/30 detector was used for TO-PRO-3. Resulting data files were analyzed using FlowJo software. 15 µm polypropylene calibration beads (PHCCBEADS, Millipore) were used to standardize cytometer settings between runs (5′000 events) and dead cells were excluded from analysis. RNA isolation, cDNA synthesis and real time PCR. Total RNA was isolated from monolayer cultures using QIAzol Lysis Reagent (Qiagen, Hombrechtikon, Switzerland) and Direct-zol RNA MiniPrep (Zymo, LucernaChem), according to the manufacturer's instructions. cDNA was prepared using VILO cDNA Synthesis Kit (Invitrogen) in a reaction volume of 10 µL following kit instruction, and diluted 1:10 with ultrapure water.
DNA isolation and determination of telomere length. Telomere length was determined by RT-PCR, as previously described 54 . Briefly, genomic DNA (from 2 × 10 5 cells) of MSC was isolated using the appropriate DNA purification kit 55 : Gentra PureGene Cell Kit (Qiagen). gDNA samples were quantified and run in triplicate in a 96-well plate in a quantitative PCR system (20 ng gDNA/well). The sequences of the both telomere and 36B4 primers are listed in Table 2. The real-time PCR program consisted of initial denaturation at 95 °C for 5 min, followed by 32 PCR cycles at 95 °C for 10 sec and 60 °C for 30 sec. Melting curve analysis was carried out for each reaction and standard curves were fitted for both telomere DNA and 36B4 DNA. For each telomere primer set, values were calculated as a ratio of telomere DNA to 36B4 DNA. In order to increase accuracy, telomere length was expressed as the average of values from the two primer sets.
Osteogenic differentiation. MSCs in monolayer at a density of 5 × 10 3 cells/cm 2 were differentiated in STEMPRO ® Osteogenesis Differentiation Kit (Gibco) for three weeks. Calcium content was determined using the Calcium CPC LiquiColor test kit (StanBio, Schwetzingen, Germany). Cells were washed with PBS, incubated with 0.5 N HCl for 30 minutes at room temperature and then with O-Cresolphthalein complexone in alkaline solution. Calcium concentration was measured (absorbance at 405 nm) and quantified with standards.
Statistical analysis. Cell volume, MSC differentiation potential markers, cytokine concentrations, telomere length and gene expression data are shown as mean values, while the data acquired at the flow cytometer is presented as geometric mean. Analysis by flow cytometry was repeated three times and results are presented as the average of three independent experiments. Kendall's tau-b correlation -a non-parametric measure -was used to assess all correlations and p < 0.05 was considered significant. Correlation coefficient (τ b ) above 0.75 was considered a high degree of correlation; correlation coefficient between 0.50 and 0.74, moderate correlation, while correlation coefficient range between 0.25 and 0.49, low degree of correlation. Absence of correlation was defined as τ b lower than 0.24. In the box plots, a non-parametric test (Mann-Whitney U-test) was used to determine the significance levels. Data analysis was performed with SPSS 24.0 for Windows (SPSS Inc.).

Data Availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.