Autofluorescence of NADH is a new biomarker for sorting and characterizing cancer stem cells in human glioma

Background The existing cell surface markers used for sorting glioma stem cells (GSCs) have obvious limitations, such as vulnerability to the enzymatic digestion and time-consuming labeling procedure. Reduced nicotinamide adenine dinucleotide (NADH) as a cellular metabolite with property of autofluorescence has the potential to be used as a new biomarker for sorting GSCs. Methods A method for sorting GSCs was established according to the properties of the autofluorescence of NADH. Then, the NADHhigh and NADHlow subpopulations were sorted. The stem-like properties of the subpopulations were evaluated by qRT-PCR, western blot analyses, limiting dilution assay, cell viability assay, bioluminescence imaging, and immunofluorescence analysis in vitro and in vivo. The relationship between CD133+/CD15+ cells and NADHhigh subpopulation was also assessed. Results NADHhigh cells expressed higher stem-related genes, formed more tumor spheres, and harbored stronger pluripotency in vitro and higher tumorigenicity in vivo, compared to NADHlow subpopulation. NADHhigh glioma cells had the similar stemness with CD133+ or CD15+ GSCs, but the three subpopulations less overlaid each other. Also, NADHhigh glioma cells were more invasive and more resistant to chemotherapeutic drug temozolomide (TMZ) than NADHlow cells. In addition, the autofluorescence of NADH might be an appropriate marker to sort cancer stem cells (CSCs) in other cancer types, such as breast and colon cancer. Conclusion Our findings demonstrate that intracellular autofluorescence of NADH is a non-labeling, sensitive maker for isolating GSCs, even for other CSCs.


Background
Glioma stem cells (GSCs) are believed to be responsible for tumor initiation, progression, chemo-and radioresistance, and recurrence of gliomas [1][2][3][4]. The identification and isolation of GSCs are crucial for a better understanding of their properties and developing GSCtargeting therapies. GSCs are usually identified and isolated from primary tumors or glioma cell lines by fluorescence-activated cell sorting (FACS) based on the cell surface makers, such as CD133 and CD15. Early studies reported that only 100 CD133-positive cells of glioma could produce a phenocopy of parent tumor in NOD-SCID mice, whereas 10 5 CD133-negative cells could not [1,5]. CD15 has also been considered as another reliable surface marker for isolating GSCs [6]. However, recent studies indicated that CD133-or CD15-negative glioma cells also possessed some GSC characteristics [6][7][8]. It is unclear whether partial CD133/CD15-negative cells have the properties of CSCs per se or partial CD133/CD15-negative GSCs are derived from CD133/CD15-positive subpopulation missing the markers by enzymatic digestion [9,10]. In addition, the antibodies of CD133/CD15 are expensive and the labeling process is time-consuming. Therefore, it is necessary to find alternative strategies, which are more specific, simple, and economic for the isolation of GSCs.
Energy metabolism is involved in the self-renewal, reprogramming, and differentiation of regular stem cells and cancer stem cells (CSCs) [11,12]. Reduced nicotinamide adenine dinucleotide (NADH) is a key carrier of electrons in cellular energy metabolism. It possesses a property of autofluorescence with an excitation wavelength at 340 ± 30 nm and an emission wavelength within the 460 ± 50 nm range [13,14], and has been used as an important intracellular autofluorescence component to non-invasively monitor and analyze metabolic activity of living cells and tissues [15,16]. Recently, NADH fluorescence intensity and fluorescence lifetime of bound and free NADH have been used to distinguish stem cells from their differentiated progeny [17][18][19]. Besides, NADH has been used to screen or monitor GSC metabolic state by using fluorescence lifetime microscopy (FLIM) [20]. However, the usability of NADH autofluorescence in the isolation and purification of GSCs by FACS has not been evaluated.
In the present study, we applied the autofluorescence of NADH as a non-labeling marker to isolate GSCs by FACS. Compared to NADH low subpopulation, NADHhigh subpopulation exhibited higher stem-like properties, including abilities of self-renewal, multilineage differentiation, and tumorigenesis, as well as higher invasive ability and resistance to chemotherapeutic temozolomide (TMZ). Besides, NADH high as a biomarker could be used to isolate breast and colon CSCs. Therefore, NADH is a suitable biomarker for the isolation of GSCs or other CSCs.

Materials and methods
Human glioma specimens and the preparation of single cell suspension A total of 13 fresh surgical glioma specimens were collected from patients enrolled in the Southwest Hospital, Third Military Medical University, Chongqing, China, after signing an informed consent from patients or their guardian. All patients had not received chemoradiotherapy before surgery. The histopathological grading was in accordance with the World Health Organization (WHO) classification (2016). The clinicopathologic information of these patients is summarized in Additional file 1: Table S1. This study was approved by the Ethics Committee of Southwest Hospital.
To prepare the single cell suspension, fresh surgical glioma tissues were collected and cut into small pieces immediately, and then, glioma cells were isolated using the Papain Dissociation System (Worthington Biochemical, Lakewood, NJ, USA) as previously reported [21,22] and suspended in PBS at 1-5 × 10 6 cells/mL.

FACS analysis and cell sorting
The cultured glioma cells were digested by trypsin or accutase and resuspended with PBS. The fresh glioma specimens were transferred to laboratory on ice in half hour after surgery, then washed and enzymatically dissociated into single cells and resuspended in PBS. The staining procedures for CD133 and CD15 markers were performed as previously described [6,8]. The labeling antibodies were anti-CD133-APC antibody (Clone REA816; Miltenyi Biotec, Germany) and anti-CD15-FITC antibody (Biolegend, USA) with REA Control (S)-APC (Miltenyi Biotec, Germany) and FITC Mouse IgM (Biolegend, USA) as controls, respectively.
The FACS analysis and cell sorting were performed on BD FACS Aria II cytometer (USA) or Beckman moflo XDP (USA). For analyzing and sorting with NADH autofluorescence intensity as a marker, an excitation wavelength of 375 nm or 355 nm and an emission wavelength of 450/50BP filter were used. For analyzing and sorting with CD133 and CD15 as markers, labeled cells were analyzed and sorted with corresponding excitation and emission wavelengths of the fluorochrome. All data were analyzed with BD FACSDiva software version 8 or Beckman moflo XDP submmit 5.2.

Limiting dilution assay
Limiting dilution assay was performed as previously described [24]. Briefly, serial twofold dilutions (from 40 to 0 cells) of different glioma, breast cancer, and colon cells were seeded into ultra-low adhesion 96-well plates (10 wells per dilution) (Costar, USA) and cultured in tumorsphere medium. After incubation for 2 weeks, wells without spheres (log2, Y-axis) were counted and plotted against the number of cells plated per well (X-axis) to calculate the sphere formation efficiency.

RNA preparation and qRT-PCR
Total RNAs from sorted cells by FACS were extracted with RNA extracting Kit (Fastagen, China) according to the

Immunofluorescence analysis
For induction of differentiation, NADH high cells were cultured in DMEM with 10% FBS for 7 days. The NADH high cells cultured in same conditions within 6 h were used as controls. Both differentiated and control cells were fixed in 4% paraformaldehyde for 30 min, washed three times with PBS at room temperature, and incubated with blocking buffer containing 10% normal goat serum and 0.3% Triton. The samples were incubated with primary antibodies anti-Sox2 (#3579, 1:400, CST), anti-Nestin (#33475, 1:400, CST), and anti-GFAP (#12389, 1:400, CST) overnight at 4°C. Hoechst 33342 was used to counterstain the cell nuclei. After washing with PBS, the samples were mounted with Immuno-Mount™ (Thermo Scientific, USA) and then examined on a LEICA TCS-SP5 confocal microscope (× 63 objective).

Xenograft in NOD-SCID mice and bioluminescence imaging
The animal study was performed in accordance with the protocol approved by the Institutional Animal Care and Use Committee of Southwest Hospital, Third Military Medical University (TMMU). NOD/SCID female mice (5 weeks old) were purchased from the Laboratory Animal Center of TMMU. Different treated GBM cells were washed and resuspended in PBS and mixed with Matrigel (1:1, BD Biosciences), then subcutaneously injected into NOD/SCID mice at 4 × 10 3 , 4 × 10 4 , and 4 × 10 5 cells (100 μL/site) with the left flank as the test group and right flank as the control group. Tumor growth was monitored by bioluminescence imaging using In Vivo Imaging System (IVIS) Spectrum (Perkin Elmer, USA) and Living Image Software for IVIS (Perkin Elmer). At the end of 6 weeks after the injection, the mice were killed. Xenograft tumors were removed and weighted.

Transwell invasion analysis
Glioma cells were seeded into the upper chambers (Millipore, 8.0 μm, 24 well) that were coated with 15 μL/ well of Matrigel in advance (Corning, USA) at the density of 3 × 10 4 cells/well in 200 μL of serum-free DMEM, and then, the upper chambers were placed in a 24-well plate added with 600 μL/well DMEM supplemented with 10% FBS. After incubation for 24 h, the cells were fixed with 4% paraformaldehyde followed by crystal violet staining. Non-invading cells were removed with a cotton swab, and the images of stained cells were collected by microscope (Olympus, Japan).

Statistical analysis
All experiments were performed at least three times. Statistical analysis was performed by using SPSS statistical software (SPSS16.0, Chicago, CA, USA) and Graph-Pad Prism 6 software (GraphPad, La Jolla, CA, USA). The unpaired two-group comparison and multiple comparisons were made with Student's t test or one-way ANOVA, respectively. Data were presented as the mean ± SD. Statistical significance was set at *p < 0.05, **p < 0.01, and ***p < 0.001.

NADH high and NADH low subpopulations can be sorted from glioma cells by FACS in vitro
By using flow cytometry, we firstly examined the autofluorescence intensity of NADH in 13 fresh glioma tissues, including 4 WHO grade II, 3 grade III, and 6 grade IV. The autofluorescence intensity of NADH was increased with WHO grades (grade IV > grade III > grade (See figure on previous page.) Fig. 1 NADH high and NADH low glioma cell subpopulations can be sorted according to their intensity of NADH autofluorescence. a The intensity of NADH autofluorescence increased with WHO grades in primary glioma cells. Also, the intensity of NADH autofluorescence in patients within same grade II (n = 4) or III (n = 3) was similar, but major difference in grade IV (n = 6) patients was observed. b Glioma cells with the highest top 10% and lowest bottom 10% intensity of NADH autofluorescence were defined as NADH high and NADH low , respectively. c The intensity of NADH autofluorescence in sorted NADH high and NADH low glioma cells was verified by confocal microscopy. All data are presented as the means ± SD. *p < 0.05, **p < 0.01 (n = 3 independent experiments) II); in low-grade gliomas (grades II and III), the autofluorescence intensity of NADH was similar between the samples, but large difference between samples was observed in grade IV (Fig. 1a, Additional file 1: Figure S1). According to previous reports [26,27], we defined the highest top 10% intensity as high autofluorescence of NADH (NADH high ) and defined the lowest bottom 10% intensity as low autofluorescence of NADH (NADH low ). Accordingly, we sorted the subpopulations with top 10% and bottom 10% intensity of NADH autofluorescence from GBM1 and LN229 cells (Fig. 1b). To confirm the autofluorescence intensity of NADH in both NADH high and NADH low subpopulations, we examined the intensity of NADH autofluorescence with confocal analysis. The cells with top 10% intensity of NADH showed strong autofluorescence intensity, while the cells with bottom 10% intensity of NADH had weak fluorescence signal (Fig. 1c). These results indicate that NADH high and NADH low subsets existed in glioma cells and could be promptly isolated by FACS.

NADH high glioma cells exhibit GSC traits in vitro
To evaluate the stem-related properties of NADH high and NADH low glioma cells in vitro, we first compared the expression of stem-related genes in both subpopulations. Compared to NADH low subpopulation, NADH high glioma cells highly expressed stem-related genes Nanog, Oct-4, Oligo2, and Sox2 at both mRNA and protein levels in GBM1 and LN229 cells (Fig. 2a). Then, the tumorsphere formation of NADH high and NADH low cells was measured through a limiting dilution analysis. In comparison with NADH low glioma cells, NADH high cells showed higher rate of tumorsphere formation in both GBM1 and LN229 cells (p < 0.01 for both) (Fig. 2b). Moreover, the average diameter of the tumorspheres derived from NADH high cells was about twice as much as that derived from NADH low cells in GBM1 and LN229 cells (p < 0.01 for both) (Fig. 2c). Previous studies showed that GSCs harbored multipotency to differentiate into neurons, astrocytes, and oligodendrocytes, and stem cell markers disappeared with the differentiation [28,29]. Hence, we evaluated whether the NADH high subpopulation had multiple differentiation potential by a differentiation assay. As expected, the differentiated NADH high cells almost lost not only autofluorescence of NADH but also neural stem/progenitor markers Sox2 and Nestin and re-expressed astroglial marker GFAP (Fig. 2d). Thus, these data strongly indicate that NADH high glioma cells have the characteristics of GSCs in vitro.

NADH high glioma cells show high tumorigenicity in vivo
The stem-related properties of NADH high and NADH low glioma cells in vivo were further evaluated by xenograft experiment in NOD-SCID mice. Bioluminescent analyses showed that the tumor size derived from NADHhigh subpopulation was significantly larger than that derived from NADH low subpopulation in LN229 cells at 28 days after implantation (Fig. 3a). As shown in Fig. 3b and Additional file 1: Table S3, the tumor incidence rate of NADH high cells was higher than that of NADH low cells. The weight of tumors derived from NADH high cells was heavier than that derived from NADH low cells (Fig. 3b). H&E staining confirmed the glioma origin of tumors, and IHC showed that the tumors derived from NADH high cells exhibited higher Ki-67 and Sox2 expression than those derived from NADH low cells (Fig. 3c). These results indicate that NADH high glioma cells have high tumorigenicity in vivo. NADH high glioma subpopulation possesses similar stemlike properties with CD133 + or CD15 + cells, but only partially overlaps with them Since CD133 and CD15 are usually used as makers to enrich GSCs by FACS, the relationship between NADHhigh , CD133 + , and CD15 + subpopulations was assessed. We first measured the proportions of CD133 + and CD15 + cells in GBM1, GBM2, T98G, and LN229 cells and found that the percentages were 0.73 ± 0.04%, 0.47 ± 0.04%, 1.37 ± 0.22%, and 0.53 ± 0.04%, and 0.50 ± 0.07%, 0.1%, 4.77 ± 0.24%, and 0.13 ± 0.09%, respectively (Additional file 1: Figure S2 and Figure S3, Table 1), which was consistent with previous reports [30][31][32]. We then compared the proportions of CD133 + and CD15 + cells in NADH high and NADH low subpopulations in those glioma cell lines. The percentages of CD133 + cells were elevated about two times in NADH high subpopulation, but no obvious change in NADH low subpopulations of those cell lines (Additional file 1: Figure S2, Additional file 1: Table S1). The proportion changes of CD15 + cells in NADH high and NADH low subpopulations were similar to those of CD133 + cells (Additional file 1: (See figure on previous page.) Fig. 2 NADH high glioma cells exhibit characteristics of GSCs in vitro. a qRT-RCR and western blotting analyses showed higher expression levels of stem-related genes in NADH high cells than in NADH low cells in GBM1 and LN229 cell lines. b, c Limiting dilution assays showed higher tumorsphere formation rates and longer average diameter of tumorsphere in NADH high cells compared to NADH low cells in GBM1 and LN229 cell lines. d Cultured in medium supplemented with 10% FBS for 7 days, NADH high cells markedly reduced the autofluorescence of NADH and the expression of neural stem/progenitor markers Sox2 and Nestin, but re-expressed high astroglial marker GFAP. Scale bar = 50 μm. All data are presented as the means ± SD. *p < 0.05, **p < 0.01, ***p < 0.001 (n = 3 independent experiments)  Table S1). The proportion of NADH high cells in CD133 + and CD15 + subpopulations was also analyzed. The percentages of NADH high cells ranged from 20.1 to 63.7% in CD133 + subpopulation and from 10 to 86.2% in CD15 + subpopulation in the four cell lines (Additional file 1: Figure S4). These results suggest that NADH high subpopulation is only partially overlapped with CD133 + or CD15 + subpopulation.

The invasion ability and temozolomide resistance of NADH high subpopulation are comparable with CD133 + and CD15 + subpopulations in glioma cells
Previous studies have demonstrated that GSCs are implicated in tumor invasiveness and chemotherapeutic resistance [33,34]. Compared with NADH low subpopulation, NADH high subpopulation had higher invasive ability in LN229 and GBM1 cells (p < 0.01 for both) (Fig. 5a). The invasive abilities between NADH high , CD133 + , and CD15 + subpopulations were comparable (p < 0.01) (Fig. 5a and Additional file 1: Figure S8). Moreover, NADH high cells were less sensitive to TMZ than NADH low cells (Fig. 5b). CD133 + and CD15 + cells were more resistant to TMZ than CD133 − and CD15 − cells (Fig. 5b), which were consistent with the previous reports [33,35]. These results suggest that NADH high subpopulation has similar malignant behaviors of invasion and chemotherapeutic resistance with CD133 + and CD15 + subpopulations.

The intensity of NADH autofluorescence can be used as a biomarker to sort other CSCs
In order to assess whether the intensity of NADH autofluorescence was suitable for isolating the CSCs in other tumors, we sorted NADH high and NADH low subpopulations from breast cancer cell line MDA-MB-231 and colorectal cancer cell line HT-29. With limiting dilution, we evaluated (See figure on previous page.) Fig. 3 NADH high glioma cells exhibit characteristics of cancer stem cells in vivo. a Bioluminescent images and quantification showed that the total flux of the tumors derived from NADH high LN229 cells (left flank) was extremely higher than that of the tumors derived from NADH low LN229 cells (right flank) at 28 days after subcutaneous implantation in NOD/SCID mice. Also, the total photon flux of the tumors was increased with implanted cell number. Signal intensity is represented as p/s/cm 2 /sr. b The images of xenograft tumors showed that NADH high LN229 cells had higher rate of tumor formation than NADH low cells (left panel). Weight statistical diagram showed that the weight of NADH high LN229 cellderived tumors was heavier than that of NADH low LN229 cell-derived tumors (right panel). c H&E staining confirmed the glioma origin of the xenograft tumors, and IHC staining showed that NADH high LN229 cell-derived tumors expressed higher expression of Sox2 and Ki67 than NADH low cells. Scale bar = 50 μm. All data are presented as the means ± SD. *p < 0.05, **p < 0.01, ***p < 0.001 (n = 3 independent experiments) the self-renewal capability between NADH high /ALDH + , NADH high /ALDH − , NADH low /ALDH + , and NADH low / ALDH − cells. Compared to NADH low /ALDH − subpopulation, NADH high /ALDH + , NADH high /ALDH − , and NADHlow /ALDH + exhibited higher ability of tumorsphere formation both in MDA-MB-231 and HT-29 (Fig. 6a). Besides, the average diameter of the tumorspheres derived from NADH high /ALDH + , NADH high /ALDH − , and NADHlow /ALDH + was larger than that from NADH low /ALDH − both in MDA-MB-231 and HT-29 (Fig. 6b). Thus, the intensity of NADH autofluorescence could be used as a biomarker to isolate CSCs from breast cancer and colorectal cancer, implying that the intensity of NADH autofluorescence might be an extensive biomarker for CSCs.

Discussion
Many endogenous ingredients of cells and tissues, such as some amino acids, collagen, elastin, NAD(P) H, flavin adenine dinucleotide (FAD), vitamins, lipids, and porphyrins, possess natural autofluorescence [36,37]. Because these Fig. 4 Both NADH high and CD133 + glioma cells possess the properties of GSCs, but are independent each other. a qRT-PCR analysis showing upregulated stemness-related transcription factor genes Nanog, Oct4, Oligo2, and Sox2 in NADH high CD133 + , NADH high CD133 − , and NADH low CD133 + subpopulations, compared to NADH low CD133 − in GBM1 and LN229. b, c Limiting dilution assay showed increased sphere formation rate and sphere average diameter of NADH high /CD133 + , NADH high /CD133 − , and NADH low /CD133 + subpopulations, compared to NADH low CD133 − cells in GBM1 and LN229 cell lines. All data are presented as the means ± SD. *p < 0.05, **p < 0.01, ***p < 0.001 (n = 3 independent experiments) endogenous autofluorescence ingredients are the metabolites of cells or tissues, their autofluorescence intensity may directly reflect the physiological and/or pathological status of cells and tissues. So far, only the autofluorescence of NAD(P) H and FAD has been widely studied, mainly to be used in monitoring alteration of metabolic profiles and cellular oxidation-reduction status [38][39][40]. Moreover, the autofluorescence of NAD(P) H and FAD has been studied in normal stem cells and CSCs. Quinn et al. reported that the quantitative metabolic imaging using the endogenous fluorescence of NADH and FAD could monitor human mesenchymal stem cell differentiation into adipogenic and osteoblastic lineages [41]. Fluorescence of free and protein-bound NADH could discriminate different differentiation stages of neuronal progenitor stem cells [42]. Buschke et al. used multiphoton flow cytometry to non- Fig. 5 The invasion ability and TMZ resistance of NADH high glioma cells are similar to the ability of CD133 + and CD15 + subpopulation in glioma cells. a The quantitative histograms of invasion showed that invasion ability was significantly increased in NADH high , CD133 + , and CD15 + LN229 and GBM1 cells. b Effect of TMZ resistance in CD133, CD15, and NADH. IC50 of TMZ in CD133 + and NADH high was higher than in CD133 − , and NADH low of GBM1 and LN229. The similar result was in CD15 of GBM1, but IC50 of TMZ in CD15 of LN229 was not different. All data are presented as the means ± SD. **p < 0.01, ***p < 0.001 (n = 3 independent experiments) invasively characterize and purify populations of intact stem cell aggregates based on NADH intensity and assessed the differentiation capacity of sorted populations [43]. Bonuccelli et al. demonstrated that NAD(P) H autofluorescence was a new metabolic biomarker for CSCs in MCF-7 breast cancer cell line and sorted high NAD(P) H autofluorescence intensity cells exhibited CSC phenotype [26]. Miranda-Lorenzo et al. used FAD autofluorescence as a novel tool to isolate and characterize epithelial CSCs, but it had obvious limitations, such as exogenous riboflavin needed to be added to enhance the sensitivity, and the experimental results varied with the concentrations of riboflavin, incubation times, and cell concentrations [44]. Therefore, in comparison with FAD, NADH autofluorescence is a more reliable and promising biomarker to be used to sort CSCs without exogenous substances to be added. In the present study, we sorted NADH high subpopulation from glioma cells and further demonstrated that this subpopulation possessed the properties of CSCs, featured with significant increase of stemness-related gene expression, tumorsphere formation, invasiveness, resistance to TMZ in vitro, and tumorigenicity in vivo.
Herein, we used a wavelength of 355 nm or 375 nm for the autofluorescence of NADH. However, under our experimental conditions, the sorted NADH high subpopulation actually also contained NADPH high cells because NADH and NADPH are spectrally identical. Nevertheless, despite the two co-enzymes exert different functions with NAD/NADH as a key determinant of cellular energy metabolism and NADP/NADPH as a central role in biosynthetic pathways and antioxidant defense, both of them may be important for stemness maintenance of CSCs. Several other studies have suggested that the concentration of NADH is higher (up to 5 times) than the NADPH in mammalian and the quantum yield of NADH is 1.25 to 2.5 times higher than that of NADPH [45]. Since NADH is the main source of the autofluorescence, we used NADH high but not NAD(P) H high subpopulation as GSCs.
CD133 and CD15 have been regarded as reliable maker for enriching GSCs. In our studies, we compared the relationship of CD133 + , CD15 + , and NADH high subpopulations and found that CD133/CD15 defines distinct cell subpopulations and both CD133 + and CD15 + cells were only partially overlapped with NADH high subpopulation in glioma cells. Thus, NADH high may define a subset of GSCs independent of CD133 + and CD15 + subsets. As for the relationship between CD133 + and Fig. 6 The intensity of NADH autofluorescence as a biomarker can be used to sort breast cancer stem cells and colon cancer stem cells. a Limiting dilution showed that compared to NADH low /ALDH − cells, NADH high /ALDH + , NADH high /ALDH − , and NADH low /ALDH + have higher sphere formation in MDA-MB-231 and HT-29. b, c The average diameter of the tumorspheres in NADH high /ALDH hig h, NADH high /ALDH low , and NADH low / ALDH high subpopulations was larger than that in NADH low /CD133 − subpopulations in MDA-MB-231 and HT-29 cells. All data are presented as the means ± SD. **p < 0.01, *p < 0.05 (n = 3 independent experiments) CD15 + cells, Son et al. reported that most CD133 + tumor cells freshly isolated from glioma specimens were CD15 + [6], but a less overlap between CD133 + and CD15 + subsets was observed in GBM1 and LN229 cells (Additional file 1: Figures S6 and S7).
As a basic metabolite, NADH is ubiquitously distributed in cells. Therefore, NADH autofluorescence could be a biomarker not only for GSCs, but also for other CSCs. Indeed, we found that NADH high /ALDH + , NADH high /ALDH − , and NADH low /ALDH + subpopulations had higher self-renewal ability than NADH low / ALDH − subpopulation in breast cancer and colon cancer cells, implying that the autofluorescence of NADH might serve as a biomarker for CSCs of these cancers.

Conclusion
Our findings demonstrate that intracellular autofluorescence of NADH is a non-labeling, sensitive maker for isolating GSCs, even for other CSCs.
Additional file 1: Table S1. The clinical features of the glioma specimens used in this study, Table S2. Primers used for qRT-PCR analyses in this study. Table S3. Tumor formation rates of xenograft implanted with different cell number of NADH high and NADH low LN229 cells. Figure S1. Intensity of NADH autofluorescence in different WHO grade glioma tissues detected by FACS. Figure S2. Representative images of flow cytometry analysis for proportion of CD133+ cells in NADHhigh and NADH low subpopulations in glioma cells. Figure S3. Representative images of flow cytometry analysis for proportion of CD15+ cells in NADH high and NADH low subpopulations in glioma cells. Figure S4. The representative flow cytometry images of the percentage of NADH high cells in CD133+/CD15+ populations. Figure S5. Both NADHhigh and CD15+ giloma cells possess the properties of CSCs, but are partially overlapped. Figure S6. The representative flow cytometry images of the relationship between CD133 + , CD15 + and NADH high populations. Figure S7. The representative flow cytometry images of the relationship between CD133 + , CD15 + and NADH high populations. Figure S8. The representative images of invasion assay for NADH high and NADH low , CD133 +/and CD15 +/subpopulations in GBM1 and LN229 cell lines. Compared to CD133 -, CD15 -NADH low subsets, CD133 + , CD15 + and NADH high cells exhibited stronger invasive ability in GBM1 and LN229 cell lines.