Discrimination of cells with stemness properties from a heterogeneous tumor cell subpopulation based on specific glycosylation pattern

The effectiveness and reliability of this new lectin combination to discriminate CSCs based on their glycan signature from whole cancer cell population were tested. For demonstrating that CSCs could be sorted with a specific MIX from A549 tumor cell line, we performed FACS analyses. Our results have shown that lectin mix conjugated to fluorescent marker only recognized around 1% of A549 cells (Fig. 1a). To confirm the specificity of lectin binding to glycan cell surfaces and eliminate false positives, we used the glycosylation inhibitor Tunicamycin, which impaired lectin labeling (Fig. 1b). Through single-cell sorting by FACS using MIX positivity, we analyzed CSCs-related transcripts, proteins and functional properties from A549 NSCLC cells. Transcriptomic analysis carried out by RT-qPCR confirmed that CSCs-related genes (Nanog, AC133, Oct4, EpCAM, Sox2) are significantly increased in MIX-positive sorted cells compared to MIX-negative cells or non-sorted cells (NSC) (Fig. 1c). Interestingly enough, compared to MIX-negative and NSC cells, CSC-related protein markers (Nanog, AC133, Oct4, EpCAM, Sox2) are significantly overexpressed in MIX-positive cells as showed in both western blotting (Fig. 1d, e) and flow cytometry (Fig. 1f, g). To determine if the proportion of MIX+ cells could change in EGFR mutated cells, we analyzed the percentage of MIX positive cells in both mutated cells lines, PC9 and H1975. Although the percentage of MIX positive cells were significantly increase with approximately 43% in H1975 expressing EGFR with activating mutations, L858R and T790M, this percentage remains close to A549 with 1% for PC9 which harbors an EGFR exon 19 deletion (Supplementary Fig. S1A). We confirmed that CSC markers, CD133, EpCAM and Sox2 are markedly elevated in MIX+ sorted cells from H1975 and PC9 compared to A549 suggesting that CSC markers are increased in EGFR-mutated cell line (Supplementary Fig. S1B, C). Previous results support this hypothesis demonstrating EGFR pathway deregulation promotes the emergence of stem like properties in non-small-cell lung cancer and is predictive of worse outcome to EGFR inhibition [15]. Since MIX+ cells expressed various CSC-related markers, we chose this one to better analyze hallmarks of CSCs. Then, we conducted clonogenicity and drug resistance tests from FACS single cell sorting in 96 wells plate. As attempted, MIX-positive A549 cells, characterized by their unique glycan signature, exhibited a significantly higher self-renewal capacity (p = 0.002) compared to negative and unsorted cells. This was significantly evident through the enhanced number and size of spheroids (p < 0.01) derived from MIX-positive-sorted cells (Fig. 1h–j). Likewise, treatment with increasing doses of Cisplatin showed significantly higher IC50 (p < 0.001) in the MIX-positive-sorted cells compared to both non-sorted and negative cells suggesting that MIX-positive cells present these resistance abilities own to CSC (Fig. 1k). Similar observations have been done in both other cancer cell lines in which the size of tumorospheres derived from MIX+ cells was significantly increased (p = 0.003) compared to those formed by MIX negative or non-sorted cells (Supplementary Fig. S1D–G). Similarly, we confirmed that cisplatin treatment induces similar effect to those observed with A549 on PC9 and H1975 cells. In the latter, cell viability measured by IC50 were significantly higher in MIX+ cells (p = 0.0016) compared to the other subpopulations, suggesting that MIX+ cells are less sensitive to Cisplatin treatment than non-sorted and MIX negative cells (Supplementary Fig. S1N, O).

Fig. 1: In Vitro characterization of LungStem kit (MIX of lectins) efficiency in a NSCLC cell line (A549).

a Representative FACS dot plots (top) showing the recognition of glycosylated patterns by the isotypic (diluent) control condition. Bottom, expression of glycosylated patterns detected by the LungSTEM kit (MIX). b Representative FACS dot plots showing the recognition of glycosylated patterns by the LungSTEM without (at left panel) and with addition of 2 µg/mL of Tunicamycin. c Representing mRNA expression levels of cancer stem cell genes (AC133, Nanog, Oct4 and Sox2) in different sorted sub-population (MIX+, MIX− and Non-sorted cells). d Western Blot shows the cancer stem cell related proteins expression in each sorted sub-population (MIX+, MIX− and non-sorted cells). e Western Blot quantifications normalized to non-sorted cells out of three replicates. f Representative flow cytometry dot plots of relative expression of AC133 and EpCAMhigh after FACS single cell sorting on LungSTEM compared with non-sorted cells. g Histogram representing EpCAMhigh and AC133+ percentages analyzed by Flow cytometry within MIX+, MIX− and Non-sorted cells. h Clonogenic capacity after FACS single cell sorting. Representative illustrations are depicted (magnification ×100). i The mean of number of colonies formed from MIX+, MIX− and non-sorted cells after FACS single sorting cells (After 60 days of incubation). j Relative colonies sizes of MIX+ and MIX− sorted cells compared with control non-sorted cells after FACS single cell sorting. The spheres’ size was monitored by taking pictures every D+7 for 80 days. k Histogram representing drug resistance to Cisplatin of MIX+, MIX− and non-sorted cells after FACS cell sorting with 1500 cells seeded per well, after 5 replicates. Results are represented as mean ± SEM, ns for not significant result, *p-value < 0.05, **p-value < 0.01, ***p-value < 0.001 using one-way ANOVA test and Student’s t test (n = 3 to n = 5 experiments).

In parallel, these findings suggest that MIX-positive cells share similar properties than CSCs sorted by CD133 (AC133). Given that a reduced percentage of MIX+ cells expressed AC133 (8.83%), we compared cancer stem cell-like properties between MIX-sorted and AC133-sorted cells in A549, H1975 and PC9 cells (Fig. 2g, h, i and Supplementary Fig. S2F, G). We observed that CD133−/MIX+ cells display respectively 40% (PC9) and 10% (H1975) of whole cell population whereas the percentage of CD133+/MIX+ cells is comparatively reduced to 20% and 1% in each cell line (Supplementary Fig S2F, G). Thus, we concluded that CD133 expression was reduced in MIX positive cells subpopulation whatever the adenocarcinoma cell lines. Therefore, we performed a Limiting Dilution Assay to challenge the self-renewal capacities between AC133++ and Mix sorted cells as this latter are composed of less AC133+ cells. Strikingly, we found that the Mix+ fraction had a significantly higher capacity to form spheroids (p = 0.0117) compared to the other fractions (Fig. 2a, b). Interestingly enough, spheroids formed by MIX+ cells exhibited a significant increase in size (p = 0.003) compared to those formed by AC133+ cells (Fig. 2c). We confirmed in both EGFR mutated cell lines, H1975 and PC9, that MIX-positive cells formed more and larger colonies compared to the CD133− fraction after analyzing AC133/MIX sorted cells, respectively p = 0.0044 and p = 0.0463 (Supplementary Fig. S2A–C). Resistance to cisplatin treatment has also been verified at various concentrations on both cell subpopulations AC133 or MIX+ sorted cells. The MIX+ A549 subpopulation cells exhibited a significantly higher IC50 (p = 0.0016) compared to the AC133+ subpopulation, indicating that MIX+ cells have greater resistance to Cisplatin treatment than AC133+ cells (Fig. 2d, e). Likewise, MIX+ cells demonstrated significantly greater resistance to treatment compared to AC133+ cells in the H1975 cell line (p = 0.0385), and a trend towards increased resistance in PC9 (p = 0.1075) (Supplementary Fig. S2D). Since stemness-related genes were found to be significantly overexpressed in both MIX+ and AC133+ sorted cells when compared to MIX−, AC133−, and unsorted cells, these results suggest that MIX+ cells present stemness properties as observed with AC133+ cells (Fig. 2f). The enrichment of CSC markers SOX2 and AC133 was notably similar in both EGFR mutated cells lines for AC133+ and MIX+ fractions contrarily to Nanog which is predominantly increased in MIX+ cells (Supplementary Fig. S2E).

Fig. 2: In Vitro characterization of MIX efficiency in lung CSCs detection compared to CD133.

a Clonogenic capacity after Limiting Dilution Assay. Representative sphere forming ability in different sorted sub-population are depicted, from AC133 (AC133+ and AC133− sorted-cells compared with Non-sorted cell (NSC) or lungSTEM MIX (MIX+ and MIX− sorted cells compared with Non-sorted cell (NSC), according to seeded cell densities (1, 10, 100, 1000 cells/well) (magnification, ×100). b Histograms represent the mean of sphere number formed after Limiting Dilution Assay with ie. NSC or MIX+ sorted cells or CD133 positive sorted cells in the condition with 1 cell/well. c Relative colony size in condition of 1 cell/well following Limiting Dilution Assay in different sorted sub-population (AC133+, MIX+ and NSC). d Response to increasing concentrations of cisplatin treatment after MACS cell sorting with AC133 or with LungSTEM MIX to assess drug resistance in MIX+, AC133+ and Non-sorted cells. e Representation of the average IC50 of Cisplatin (µM) in the different sorted subpopulations: MIX+, AC133+ and Non-sorted cells. f Analysis of CSCs genes levels (Nanog, AC133, Oct4, Sox2) in different sorted sub-population (MIX+, AC133+ and Non-sorted cells). g Representative FACS dot plots showing the basal expression of AC133 (at left), LungSTEM MIX (at middle) and coexpression (at right) in Non-sorted cells. h Representative FACS dot plots showing AC133 expression (at left), LungSTEM MIX expression (at middle) and coexpression of AC133 and MIX (at right) in Mix+ sorted cells. i Representative FACS dot plots showing AC133 expression (at left), LungSTEM expression (at middle) and coexpression of AC133 and MIX (at right) in AC133+ sorted cells. All results are represented as mean ± SEM, ns p-value indicate not significant result, *p-value < 0.05, **p-value < 0.01, ***p-value < 0.001 using one-way ANOVA test (n = 4 experiments).

Thus, our data revealed that the MIX is more selective for lung CSCs compared to the AC133 antibody-based sorting method. These results demonstrate that the MIX significantly highlight a fraction of cells with CSC hallmarks.

MIX-sorted cells express markers and functional properties of epithelial to mesenchymal transition (EMT)

Since CSCs were also endowed with epithelial-to-mesenchymal transition, we analyzed the cells ability to migrate and invade. Hence, we performed wound healing scratch and invasion tests following FACS sorting cells based on MIX labeling. We found that, fraction expressing the glycan signature recognized by the MIX showed a significantly increased migration capacity compared with both MIX− (p < 0.001, orange curve) and unsorted fractions (p < 0.05, black curve) (Fig. 3a, b). These results were confirmed in the other two mutated cell lines, H1975 and PC9, where the MIX+ fraction also showed a significantly higher migration capacity compared to other fractions (p < 0.0001) (Supplementary Fig. S3A–D).

Fig. 3: Implication of MIX+ stem like cells in the migration and invasion.

a Representative illustrations of wound healing scratch assay are depicted using Incucyte 2022 Rev1 (magnification ×10). b Wound healing scratch assay for unsorted, MIX+ and MIX− sorted cells. Percentage of wound healing was measured using Incucyte 2022 Rev1 Software. c Western Blot shows Epithelial-mesenchymal transition (EMT) related proteins levels in each sorted sub-population (MIX+, MIX− and non-sorted cells). d Western Blot quantifications normalized to non-sorted cells out of three replicates. e Representative images of spheroids’ invasion capacities after 5 days of incubation in Matrigel coated inserts, in each sorted sub-population (MIX+, MIX− and non-sorted cells). f Histogram representing the relative invasion capacity of spheroids from each sorted subpopulation (MIX+, MIX− and unsorted cells). Spheroids from each subpopulation were embedded in Matrigel. Matrigel invasion was measured by deducting the total area from the central area, using the Fiji Macro analysis program. Some results are represented as mean ± SEM, ns not significant result, *p-value < 0.05, **p-value < 0.01, ***p-value < 0.001 using one-way ANOVA test (n = 3 experiments) and others Student’s t test (n = 4 experiments).

To strengthen these findings, we performed western blotting analysis to assess the expression of EMT-specific proteins such as Snail and ZEB1. Remarkably, we observed that MIX+ sorted A549 cells exhibited a significant upregulation of these markers (p < 0.001) (Fig. 3c, d). Similar results were observed in the PC9 (p = 0.003) and H1975 (p = 0.0009) lines (Supplementary Fig. S3E, F).

Additionally, spheroids previously obtained after FACS single cell sorting based on Mix staining were seeded on the surface of a Matrigel matrix to observe cell invasion. Notably, a distinct halo of invasive cells was observed surrounding the spheroids exclusively only in the MIX+ and non-sorted cells. As observed, the cell ability to invade and migrate was significantly concentrated in the MIX+ population (p = 0.0013) (Fig. 3e, f). These observations were also noted and quantified in the H1975 (p = 0.0009) and PC9 cells p = 0.0034) (Supplementary Fig. S3G, H, I, J).

Taken together, these functional characterization results highlighted that the MIX+ subpopulation exhibits CSCs hallmarks such as specific stemness markers, clonogenicity, drug resistance as well as invasion ability. These results would suggest that in vivo, MIX+ cells would present more aggressive behavior than another cell fraction.

Tumorigenic potential of MIX+ sorted cells

To demonstrate that MIX+ cells could be related to CSCs, MIX+ tumorigenicity was analyzed in immunocompromised animal host. After FACS sorting, 45 nude mice were injected subcutaneously with different amounts of cells sorted either by MIX+, MIX− or unsorted dilutions, as shown in Fig. 4a.

Fig. 4: In Vivo tumorigenesis capacities of MIX positive cells efficiency in lung CSCs detection: tumorigenic potential of Mix+ and Mix− cells isolated from A549.

a Illustration of the process followed after FACS-sorting on the LungSTEM (MIX) and injection to different mice. MIX+ sorted cells and MIX− sorted cells were subcutaneously transplanted into nude mice. b Illustration representing the number of tumors formed (tumor > 100 mm3) by mice according to the decreasing number of injected cells from each subpopulation (MIX+ and MIX−). c Tumors growth curve representing the evolution of tumor size after 500 cells injected in each mouse. It was monitored by the size from 1 to 60 days after the cells were injected into nude mice. The tumor size was the average of the vertical and horizontal diameters (n = 5 mice for each condition). d Representative illustrations of Ki-67 staining are depicted with higher magnification (×200). e Histogram representing Ki-67 staining measured in each subpopulation (MIX+ and MIX−) using the Fiji Macro analysis program on eight sections. All results are represented as mean ± SEM, ns not significant, *p-value < 0.05, **p-value < 0.01, ***p-value < 0.001 using Student’s t test.

The number of mice with tumors larger than 100 mm3 was counted sixty days after injection (Fig. 4a). Interestingly, among the 15 mice transplanted with the MIX− sorted cell subpopulation, only 3 mice developed tumors whereas 7 mice have tumors derived from MIX+ sorted cells (Fig. 4b). Strikingly, MIX+ cell subpopulation remains the sole condition to observe initiating tumor formation from the 500 and even 50 cells injected grafts (Fig. 4b). These results suggest that the tumorigenicity of MIX+ cells are higher than those of MIX−. Since a limited number of MIX positive cells (50 cells) are required to initiate tumors, we can conclude that MIX+ cells with their stem-like features would be more aggressive.

The analysis of the tumor volume from the different xenograft supports our initial findings. Indeed, MIX+ derived tumors had an average tumor volume 6 times greater than tumors derived from other cells (p = 0.0005). MIX+ sorted cells resulted in larger tumor volumes (300 mm3) compared to MIX− sorted cells (50 mm3). Importantly, MIX+ cells achieved a mean tumor volume exceeding 100 mm3 within 40–45 days after a 500-cells injection, whereas no tumors reached this size with MIX− (Fig. 4c). These results suggest that tumors obtained after injection of the MIX+ sorted subpopulation remains more aggressive than those obtained with the MIX− sorted subpopulation.

We assessed cell proliferation rates in the tumors using Ki67 staining. Consistent with expectations, the immunohistochemical analysis of Ki67 in tumor sections from both the MIX+ and MIX− subpopulations revealed a significant increase proliferation (p = 0.0013) in the MIX+ group compared to the MIX− group (Fig. 4d, e).

Altogether, these results would confirm in vitro findings and demonstrate that the sorted MIX+ subpopulation is enriched in CSCs due to their ability to initiate tumors even with a significantly reduced grafted cells number.

Detection of NSCLC patients with MIX+ cells

The in vitro and in vivo experiments revealed that MIX-positive subpopulation exhibits CSCs features and also could be related to Lung CSCs. In these conditions, we decided to study its clinical relevance in discriminating healthy tissue from lung tumor. We analyzed whether MIX-positive could accurately detect lung CSCs directly in tissue samples by IHC. The staining with MIX was performed on 221 tumor tissues from two different cohorts (from Lyon University Hospices and AMSBIO) (Table 1). Among the stained cell TMA, 86 were scored as MIX-Low (0–30% stained cell) and 128 as MIX-High (>30% stained cell) (Fig. 5a, Table 1). Independence between MIX staining and others clinical variables were confirmed by Chi-2 test (vs Sex: p = 0.17; vs Age class: p = 0.81; vs Stage, p = 0.33).

Fig. 5: Validation of LungSTEM MIX staining efficiency in the detection of NSCLC patients.

a Representative illustrations of LungSTEM (MIX) staining (in brown), as observed by IHC are depicted (magnification, 200×). b The Receiver Operating characteristic (ROC) curve of LungSTEM (MIX) and mainly performance parameters of the classification (Accurancy, sensitivity and specificity).

First, we evaluated the specificity and sensitivity of MIX staining using 70 tumoral lung tissues (36 early stage and 34 late stage) and 63 paired non-tumoral tissues. None of non-tumoral harboring positive-cells after MIX staining, indicating an excellent specificity (100%). After application of 30%-cutoff of MIX-staining, 19 tumoral tissues can be considered as false negative, reaching a sensitivity of 72.1% (Fig. 5b). ROC curve built from results of MIX staining on non-tumor and tumor lung tissue permit to establish that AUC is acceptable (AUC = 0.72; data not shown) and confirmed a good accuracy (86%, 95% CI: 0.78–0.91; p = 5 × 10−16; Fig. 5b). These results suggest that the MIX might be likely to discriminate CSCs from healthy stem cells. (Fig. 5b, c).

NSCLC patients grading based on tumor aggressiveness and prognostic value of MIX+ staining

Secondly, we aim to assess whether the MIX-positive cells could predict patient relapse or survival prognosis. Overall survival (OS) analysis were evaluated by Kaplan–Meier curves and Cox regression models, according to MIX staining (MIX-Low and High) and others clinicopathological data, i.e. sex (men and women), age (< and ≥60 years old) and stage (early, I/II and late, III/IV). Noted that OS analysis has been performed on whole patients (n = 221) from HCL and AMSBIO cohorts (Table 1), while RFS has been conducted only on HCL cohort (n = 70), because AMSBIO failed onto available recurrence information.

Prognostic significance for OS of MIX scoring is weakly supported by survival curve (p = 0.13; Supplementary Fig. S4A and Fig. 6b) and univariate Cox model (HR: 1.3 with 95% CI 0.91–1.95, p = 0.136; Fig. 6a). Noted that sex and age have not a significant impact on survival rates (p = 0.24 and p = 0.457, respectively). However, OS analysis demonstrate that late stage (III/IV) is a poor prognosis factor (HR: 2.8 with 95% CI 1.9–4.05, p = 6.6 × 10−8; Supplementary Fig. S4A and Fig. 6b). Since multivariate analysis revealed that late stages and a high-MIX score were independent prognosis factors of patients’ outcome (Supplementary Fig. S4B), we chose to analyze prognostic value of MIX-staining on early stage and late stage, separately. Thus, survival curves highlight that high MIX staining is a bad prognosis factor on early stage of NSCLC (p = 0.016; Supplementary Fig. S5A, B) but is not informative to the prognosis of late stage (p = 0.91, data not shown). High MIX staining in early stage harbor a hazard ratio of 2.1 (95% CI: 1.13–3.83, p = 0.018, data not shown). To accurately estimate the prognosis value of MIX staining according to given stages, survival analysis (Kaplan–Meier and univariate Cox regression) according to Low or High MIX subpopulations at early (I/II) or late (III/IV) stages, were performed (Fig. 6b). MIX-High staining could be clearly considered as a poor prognosis marker only early stage. Since Kaplan Meier curves were still performed on two populations of patients combined (HCL and AMSBio), we evaluated independently each population to eliminate any bias due to the sample size. When the curves corresponding to each cohort are analyzed separately, we confirmed the MIX staining might be useful as a prognosis value regarding overall survival in early stages in comparison to all stages and late stages (Supplementary Fig. S5A–I).

Fig. 6: Prognosis value of MIX staining combined with stages on relapse-free survival and overall survival.

Relapse-free survival (RFS) from Lyon University Hospital cohort (a) and overall survival (OS) from both cohorts (b) was analyzed according to pTNM stage (early stage and late stage) of NSCLC patients and LungSTEM (MIX) staining intensity (low/high). Kaplan–Meier curves are depicted according to MIX staining combined with lung adenocarcinoma stages. P value indicated in each panel correspond to log-rank test (Mantel–Cox) performed to survival curve comparison (RFS and OS). Dashed line highlights RFS or OS median for each subgroup of patients. Risk table was indicated for each Kaplan–Meier plot to show the number of patients at risk by time, for each subgroup. Hazard ratio (HR), confidence interval of HR (CI) and p value are evaluated with Univariate Cox regression model in RFS and OS. pTNM pathology tumor–node–metastasis.

As the same way, we have conducted RFS analysis. Only sex and stage have a significant prognostic value for recurrence in NSCLC and MIX staining appeared as no informative (data not shown). Nevertheless, we have pursued our RFS analysis, according to Low or High MIX subpopulations at early (I/II) or late (III/IV) stages. We show that late stage patients with High MIX staining seemed to have a worse overall survival (p = 0.13; Fig. 6a) compared to early stage patients with low staining. Univariate Cox model confirm that high MIX staining is a bad prognosis factor for recurrence (HR: 2.1 with 95% CI: 1.04–4.14, p = 0.038; Fig. 6a).

These results highlight the potential of MIX staining as a marker of tumor aggressiveness that could complement the existing TNM classification. Interestingly, it could reflect the CSCs Burden in tumor samples and thus could be clinically valuable for predicting patient outcomes in early-stage lung adenocarcinoma and might help to treatment decisions by providing insight into relapse risk.

Multiplex immunohistochemistry analysis showed that there were no B lymphocytes (LB, in red) infiltrating the lung tumor tissue at any stages. CD8+ T cells (CD8+ T cells, in green) are significantly increased in the area in which the MIX positive cells are localized at late stages. However, CD8+ T cells significantly reduced in non-tumoral lung tissue (p = 0.0001; Supplementary Fig. S6A, B). Likewise, in early stages, CD8+ T cell infiltration was significantly decreased, despite the presence of MIX-positive cells associated with CSCs in the tumor (p = 0.0001; Supplementary Fig. S6A, B).

These results were supported by previous findings demonstrated that CSCs create a tolerogenic immune microenvironment, promoting their survival and resistance to conventional treatments such as chemotherapy [16]. Thus, despite the infiltration of numerous immune cells surrounding CSC in lung adenocarcinoma, these immune cells do not necessarily exert anti-tumor activities [17].