Creation of a RFP-GFP-LC3B lentiviral vector enabling analysis of autophagy activity in single viable cells
To facilitate measurements of the autophagosome content of individual viable cells, we created a lentivirus encoding a RFP-GFP-LC3B tandem construct (Fig. 1A). This construct allows autophagosomes to be seen as fluorescent yellow vesicles because they are positive for both GFP and RFP, and also distinguished from autolysosomes, which display a red fluorescent signal because the GFP signal is quenched in the acidic environment of the lysosomes with which the autophagosome has fused. We then used confocal fluorescence microscopy to examine the content of autophagosome foci in transduced MCF10A cells selected by FACS for their differential expression of RFP (R+ cells, autophagy-high) or GFP (G+ cells, autophagy-low). The results of these confocal measurements confirmed that the G+ cells were predominantly expressing autophagosomes, whereas the R+ cells contained more autolysosomes (Fig. 1B), in agreement with previously reported data for cells transfected with the same internal construct24,25.
We then exposed transduced, but unselected MCF10A cells to 1 µM rapamycin for three hours and reanalyzed the cells for RFP and GFP expression by FACS at the end of the exposure period (Fig. 1C). This showed rapamycin exposure increased the proportion of R+ cells, and diminished the proportion of G+ cells, compared to controls (Fig. 1D). Western blot (WB) analysis of the isolated R+ and G+ cells showed that the control G+ cells contained equivalent levels of LC3B-I and LC3B-II, whereas the R+ cells contained mostly the LC3B-II form (Fig. 1E). This increase in the ratio of LC3B-II to LC3B-I in the R+ cells as compared to the G+ cells served to demonstrate the ability of flow cytometric analysis of RFP-GFP-LC3B–transduced cells to reveal induced changes in autophagy activity at the single-cell level.
Different subsets of normal human mammary cells display similar autophagy flux activity despite different levels of ATG-related proteins
Three major types of human mammary cells; basal cells (BCs), a luminal progenitor-containing subset (LPs) and a more mature luminal cell subset (LCs) can be individually isolated in a viable state directly from enzymatically dissociated normal female reduction mammoplasty tissue by fluorescent-activated cell sorting (FACS) based on their separate or dual expression of two cell surface markers, EpCAM and α6-integrin (recognized by antibodies to CD326 and CD49f, respectively) in combination with antibodies to CD45 and CD31 to remove contaminating hematopoietic cells and endothelial cells, respectively. The remaining non-mammary, stromal cells (SCs) present can also be separately isolated using this strategy as they are negative for all of the above markers21,26,27.
In a first series of analyses, we isolated BCs, LPs, LCs and SCs at high purity (> 96%) from the same donor samples and examined their content of 4 key ATG proteins; ATG7, BECLIN1, LC3B-II and ATG4B by WB analysis. This showed consistently higher levels of both ATG7 and BECLIN1 in the LPs than in the BCs, and intermediate levels of these in the LCs and SCs, with similar trends for LC3B-II and ATG4B (Fig. 2A,B and Supplementary Fig. S1A).
Since inhibition of the autophagic maturation step mediated by fusion of the autophagosomes with lysosomes could lead to an accumulation of both LC3B-II and autophagosomes, we next asked whether the higher levels of ATG proteins in the LPs compared to the BCs might reflect a more active autophagosome generation process in LPs rather than a slower rate of autophagosome-lysosome fusion. To distinguish between these alternatives, we treated purified normal human mammary BCs and LPs in vitro for two hours with 200 nM BafilomycinA1 (BafA1), an established inhibitor of the autophagosome fusion step5. WB comparison of the LC3B-II content in the BafA1-treated cells compared to controls showed the same increased levels of LC3B-II in both the treated BCs and LPs (Fig. 2C), indicating that their different levels of LC3B-II was not explained by differences in their rates of autophagosome-lysosome fusion.
We then used the lenti-RFP-GFP-LC3B vector to examine the properties of individual BCs and LPs. FACS analysis was performed on cells maintained in vitro in supportive EGF-containing medium for four days after exposure to the vector (to allow for maximal uptake and expression of the transgene, Fig. 2D). This revealed that a high frequency of gene transfer efficiency to both subsets (~ 70%) was consistently achieved, but the levels of RFP and GFP expression in the individual cells in both were very heterogeneous. WB analyses of FACS-isolated R+ cells from both BCs and LPs showed a higher content of LC3B-II relative to LC3B-I than in the matching G+ cells isolated from the same samples (Fig. 2E). WB analysis also showed P62 levels to be lower in the isolated R+ cells compared to the G+ cells, confirming an increased rate of autophagy-mediated protein degradation in the R+ cells (Fig. 2E). Importantly, R+ cells from both the BC and LP subsets did not contain higher levels of ATG7 or ATG4B, again consistent with primary human mammary R+ and G+ cells being equally autophagy-competent (Fig. 2C and Supplementary Fig. S1B). On the other hand, the fact that BC and LP became similar in their content of ATG7 and ATG4B after 4 days in vitro (Fig. 2E) suggests that the overall levels of these ATG proteins are differentially affected by the culture conditions used, compared to those operative in cells isolated directly from normal human breast tissue.
In addition, because the autophagy process is known to be a dynamic one, it was of interest to examine its stability in the mammary cells maintained for a more prolonged time in vitro. Analysis of the progeny of R+ BCs or LPs maintained for seven days in vitro under the same culture conditions showed they remained R+ (Supplementary Fig. S2). In contrast, however, initially G+ BCs or LPs produced similar numbers of G+ and R+ progeny (Supplementary Fig. S2).
Human mammary cells with progenitor activity require autophagy activity despite variable baseline levels
To determine whether autophagy is important to the proliferative potential of normal human mammary cells we first transduced FACS-purified BCs and LPs with the RFP-GFP-LC3B vector, and then, four days later, isolated the R+ or G+ fractions of each and assayed them for clonogenic activity in standard 9-day CFC assays in vitro. The results showed that the CFCs in both the BC and LP subsets were also highly variable in their distribution in their levels of LC3B expression, but overall, appeared slightly enriched in the G+ fractions, more prominently in the BCs (P = 0.04, Fig. 3A).
We then asked whether a similar result would be obtained in the BCs and LPs that are regenerated in the bilayered mammary epithelial structures that normal BCs produce in collagen gels transplanted under the kidney capsule of immunodeficient mice28,29. Accordingly, mice were transplanted with BCs transduced with the lenti-RFP-GFP-LC3B virus and four weeks later, gels were harvested, single cell suspensions prepared. The R+ and G+ BCs and LPs in them were then isolated by FACS and the different fractions plated in vitro in the same type of CFC assays. In this case, the results showed that the CFC activity (frequency) in the BCs was higher in the R+ subset (higher LC3B-II), but in the LPs, was again not significantly different in the R+ and G+ fractions (Fig. 3B).
To examine whether BC and/or LP progenitor activities might be dependent on functional autophagy capacity, we then compared the yield of colonies obtained in CFC assays to which 200 nM BafA1 or 10 µM chloroquine (CQ) or vehicle was added. In both the BafA1– and chloroquine-treated BCs and LPs, blocked fusion of the lysosomes with the autophagosomes was evident as shown by an increase in LC3B-II (Supplementary Fig. S3A). Interestingly, however, both of these autophagy inhibitors significantly and selectively decreased colony yields from the BCs (by 40–50%, P < 0.05), with no detectable effect on colony yields from the LPs (Fig. 3C).
Given the finding that ATG7 levels were similar in the cells generated from BCs and LPs cultured under similar conditions, but at non-limiting cell concentrations (Fig. 2E), and the specific requirement of ATG7 for autophagic function via its role in converting LC3-I to LC3-II30,31, we also compared the effect of specifically inhibiting the expression of ATG7 on BC and LP clonogenic activity. For this, we transduced freshly isolated BCs and LPs with either of 2 inducible shATG7-puromycin-encoding or a control lentiviral vector and then selected the transduced cells for 2 days in puromycin prior to plating the selected cells in CFC assays. In presence of doxycycline, the BCs transduced with the shATG7 vector showed a 50–60% reduction in the number (and size) of colonies produced compared to controls, whereas the clonogenic activity of the similarly transduced LPs was not affected (Fig. 3D and Supplementary Fig. S3B-C). Taken together, these results show that the clonogenic activity of the phenotypically and functionally distinct BCs and LPs of the normal human mammary gland contain different levels of ATG proteins and respond differently to autophagy inhibition.
Autophagy levels distinguish human mammary cells with heterogeneous tumor-initiating activities
We next sought to determine whether changes in autophagy components or activity occur during the initiation of human mammary cell transformation. For this, we analyzed the cells in tumors produced rapidly from normal human primary mammary cells transduced with a KRASG12D-encoding vector and then transplanted into female immunodeficient mice22. In an initial set of experiments, we examined the effect of a two-hour exposure of KRASG12D– or control vector-transduced BCs and LPs to 200 nM BafA1 on the expression of LC3BI and LC3B-II in the cells present after another three days in vitro. WB analysis of the cultured KRASG12D-transduced BCs or LPs exposed to 200 nM BafA1 showed an equivalently increased level of expression of LC3BI and LC3B-II compared to the corresponding cultured subsets of control cells (Fig. 4A).
We then used IHC to determine how LC3B levels might be altered in the early transformed progeny of KRASG12D-transduced cells that are initiating tumor formation two weeks after their transplantation into immunodeficient female mice22. Comparison of the LC3B levels in the normal tissue from which the transduced cells had originally been isolated showed the cells in the nascent tumors contained less LC3B (Fig. 4B and Supplementary Fig. S4A).
We then designed an experiment to investigate whether the initial autophagy status of the cells would affect their susceptibility to KRASG12D-initiated transformation. For this, we first transduced the cells with the lenti-RFP-GFP-LC3B virus, and then four days later, separately isolated their R+ and G+ derivatives. Equivalent numbers of each of these four cell populations (R+ BCs and LPs, and G+ BCs and LPs) were then transduced independently with the KRASG12D-virus and the cells then immediately transplanted into mice. Subsequent bioluminescent tracking of their rates of tumor formation showed the progeny of the G+ cells were already growing more rapidly than the R+ cells by 2 weeks post-transplant (Fig. 4C), despite a similar efficiency of KRASG12D transduction of the initial G+ and R+ cells (Supplementary Fig. S4B). The more rapid growth of the KRASG12D-transduced G+ cells was confirmed macroscopically using an EVOS fluorescent imaging microscope to size the tumors harvested another 2 weeks later; i.e. 4 weeks post-transplant (Supplementary Figure S4C).
To determine if the initial growth advantage displayed by the KRASG12D-transduced cells with low autophagy would be perpetuated, 4-week primary tumors initiated from G + cells were dissociated, again sorted into R+ and G+ phenotypes and the same numbers of each were then transplanted into secondary female mice (5000–30,000/mouse). However, in this case, the rate of secondary tumor growth proved to be the same regardless of the autophagy activity (R+ or G+ phenotype) of the primary tumor cells from which the secondary tumors were generated (Fig. 4D). Thus the greater primary tumorigenic activity of the KRASG12D-transduced G + cells isolated directly from human mammary tissue (Fig. 4C) was not sustained.
To determine whether this correlation would extend beyond the KRASG12D oncogenic stimulus, we applied the same strategy to MCF10A-BMPR1B+ cells. These cells have been selected for BMPR1B expression and were then further exposed for 8 weeks to a combination of BMP2 and IL6. MCF10A-BMPR1B+ cells are tumorigenic in immunocompromised mice and able to form colonies efficiently in soft agar in contrast to parental MCF10A cells that have neither of these properties32. An initial comparison of the transcriptional profiles of the parental MCF10A cells with the MCF10A-BMPR1B+ showed the latter display increased autophagy regulatory pathways (Fig. 5A), suggesting the pathways activated by BMP2 and IL6 might be involved in the early steps of transformation. Stable RFP-GFP-LC3B MCF10A-BMPR1B+ cells were then generated, and R+ or G+ fractions isolated by FACS (Fig. 5B). Similar to the response of primary human mammary cells transduced with KRASG12, the G+ MCF10A-BMPR1B+ BMP2-IL6-transduced cells displayed an increased capacity to form soft agar colonies compared to the R+ cells (Fig. 5C), suggesting that a higher propensity for low autophagy cells to become transformed regardless of the mechanism(s) inducing their transformation.