A phenotype-based cell culture model of melanoma cells that persist as residual disease after therapies leading to recurrence

TL;DR

This paper introduces a cell culture model to study melanoma cells that survive therapies by entering deep quiescence, leading to recurrence.

Contribution

A novel phenotype-based cell culture model to study therapy-resistant melanoma cells in deep quiescence.

Findings

Glutamine deprivation selects rare melanoma cells capable of deep quiescence and long-term survival.

Cells selected under this model show increased resistance to paclitaxel compared to parental lines.

The approach works with both human and mouse melanoma cell lines, indicating broad applicability.

Abstract

A crucial adaptability trait of stem-like melanoma cells that persist under all selection pressures, including therapies, is their ability to survive in deep quiescence. This ability coupled with their intrinsic ability to proliferate leads to recurrence. Here we describe an approach to modeling this trait in cell culture. A lack of glutamine proved to be a selection pressure for the highly metastatic human melanoma cell line A375SM, killing more than 99% of cells and selecting rare cells based on their ability to survive in deep quiescence. After 4 weeks, cells gradually exited quiescence and proliferated indefinitely. Interestingly, by not providing fresh glutamine-free medium at this stage, we could select rare cells that persist in deep quiescence as single cells. Alternatively, we could model deeper quiescence lasting longer than 4 weeks by increasing the severity of the selection pressure using dialyzed serum in medium. The cells selected in this manner were much more resistant to paclitaxel than was their parental cell line. We obtained similar results with the highly metastatic mouse melanoma cell line B16-BL6. Thus, our phenotype-based approach is suitable for modeling abnormal deep quiescence in melanoma that is responsible for therapy resistance, disease progression, and recurrence/metastasis.

Figures

a](#F3)). Specifically, cells did not exit quiescence by week 7 in stringent Gln-free medium, whereas they exited quiescence after week 4 in regular Gln-free medium. One important question is whether lack of proliferation in stringent Gln-free medium results from irreversible severe damage to cells. To answer this question, after 28 days of incubation of A375SM cells in stringent Gln-free medium, we shifted the cells to complete medium with Gln and regular FBS. We found that upon removal of a severe bottleneck in this manner, cells exited quiescence and proliferated vigorously ([Fig. 3b](#F3))

a](#F4)). Next, to determine whether the cells surviving after 6 days of treatment could grow into colonies, we allowed paclitaxel-treated cells to recover in a medium without paclitaxel and grow into colonies for 9 days. We found that paclitaxel-treated A375SM-MA cells yielded significantly more colonies than did similarly treated parental cells, indicating a significantly higher chemotherapy resistance of A375SM-MA cells than parental cell line ([Fig. 4b](#F4)).

a](#F6)). Significantly, like the results obtained with A375SM cells ([Fig. 3](#F3)), B16-BL6 cells surviving in quiescence failed to exit it for up to 7 weeks. When we shifted the cells to complete medium for a week, cells exited quiescence and proliferated, indicating that the surviving cells were not irreversibly growth-arrested ([Fig. 6b](#F6)). Of note was a high degree of inter-colony and intra-colony heterogeneity of the cells.

a](#F7). To compare paclitaxel’s effect in another manner (by comparing colony-forming ability/proliferation potential of cells that survived paclitaxel treatment), we shifted the paclitaxel-treated cells to drug-free medium for 9 days and then stained the dishes containing these cells with crystal violet ([Fig. 7b](#F7)). This provided an estimate of the proliferative ability of the cells remaining after 6 days of treatment with paclitaxel. The number of B16-BL6-MA cells surviving treatment with 80 nM paclitaxel, likely in quiescence and then growing into colonies in drug-free medium, was dra

Full text

Introduction

Targeted therapies and immune checkpoint therapies have significantly improved clinical management of melanoma cases in recent years^1,2^. However, further improvements are still urgently needed for decreasing the risk of recurrence and metastasis. Residual disease after the initial treatment of melanoma often evolves further to cause recurrence and/or metastasis. Therefore, developing and improving therapeutic strategies to prevent and delay recurrence are critical. Residual disease requires unique considerations regarding potential therapies that are different from those applicable to the initial treatment or even the metastasis stage. Because of the dormant nature of disease, administration of therapies for long periods may be needed for preventing or delaying recurrence. Therefore, they must be safe and effective against evolving heterogenous disease. These factors make therapy development for preventing cancer recurrence very challenging. Furthermore, because of high failure rates, clinical trials are often prohibitively expensive.

Effective preclinical modeling of residual disease for testing therapies may improve the success rate of clinical trials in preventing recurrence and/or metastasis. To that end, how do we model the cancer cells responsible for recurrence? The cancer cells that must be modeled should be highly aberrant as well as highly adaptable, a combination that enables them to cause recurrence. To bring these two traits together, we can 1) rely on the disease context from which cancer cell lines are derived (for high cancer cell abnormality) and 2) impose severe, prolonged metabolic challenge as a bottleneck (for selecting a high level of fitness). A crucial adaptability trait of such cells is their ability to opportunistically switch to quiescence or proliferation depending on their environment^3^. A cell culture system may be ideal for modeling such cells, allowing for direct observation of cells capable of surviving in deep quiescence. Equally important, a cell culture system offers the convenience of enabling evaluation of low-dose therapies over long periods, making such evaluations more predictive of response in the clinic than other preclinical models.

When cancer cells are isolated from primary tumors or metastases and initially placed in a cell culture medium for generating a cell line, only a small subpopulation of cells succeed in proliferating. The cells that do so are no longer under body-like bottlenecks/selection pressures. We believe that severe selection pressures serve as niches for an adaptable cancer cell state driving cancer evolution in the body. Therefore, even when a cell line is established from an incurable metastatic cancer, most cells growing in an artificially rich medium (lacking body-like bottlenecks) are not adaptable. Consequently, they can easily be killed with current therapies targeting the proliferative cell state. The key to modeling an adaptable cancer cell state is to apply a severe bottleneck to select evolutionarily fit rare cells (~0.01% of the population). Such cells must also survive in deep quiescence, which is a defining feature of residual disease. Quiescence can be monitored under a microscope over time before surviving cells overcome the bottleneck and begin proliferating. Finally, we consider exit from quiescence and cell proliferation as essential features of a model representing a cancer with high risk of recurrence. This straightforward design allows for modeling of the adaptable cancer cell state in a usable system for testing therapies. Although the bottlenecks that can be applied in cell culture are different from those operating in the body, what is important is that the same cancer cell phenotype (deep quiescence) modeled in cell culture helps cancer cells survive and evolve further under all bottlenecks and therapies.

We serendipitously discovered the feasibility of modeling adaptable cancer cell state featuring abnormal deep quiescence in cell culture while studying highly aggressive triple-negative inflammatory breast cancer^4–7^. Because the concept of abnormal quiescence as a defining feature of cancer cell adaptability is applicable to all resistant cancers, in the present study, we investigated whether this approach would work with melanoma. One reason for choosing melanoma from among the resistant cancers is that a lack of glutamine (Gln) deep in melanoma lesions is associated with a resistant cancer cell state due to an altered epigenome^8^. This suggested to us that a lack of Gln in cell culture could serve as a body-like bottleneck for selecting resistant melanoma cells of a desired phenotype (i.e., survivability in quiescence followed by further evolution). Using two highly metastatic melanoma cell lines of human and mouse origin, we provide evidence supporting the feasibility of this cell culture approach for selecting highly resistant rare melanoma cells that survive in deep quiescence.

Results and Discussion

Survival of rare melanoma cells in deep quiescence in Gln-deficient culture medium

The types of cancer cells that are responsible for recurrence are highly abnormal and adaptable and thus capable of persisting and evolving. Optimal modeling of deep intrinsic resistance depends on both evolutionary fitness of the cancer cells and the nature and severity of the bottleneck for selecting the fittest cells. To optimally model high cancer cell abnormality, we chose the highly aggressive human melanoma cell line A375SM^9^. As an added benefit, this cell line also harbors clinically actionable active BRAF (V600E) mutation. Modeling a realistic body-like bottleneck is a major challenge in cell culture. In the body, cancer evolution results in selection of cancer cells that can survive a variety of challenges, such as lack of nutrients, and many additional challenges designed to enforce a code of multicellularity^10,11^. A lack of Gln may be an easy-to-implement ideal bottleneck in cell culture for some cell lines because proliferative cancer cells are addicted to Gln^12^. In this regard, the fact that Gln is a nonessential amino acid may also be advantageous, increasing the likelihood that a severe bottleneck can be applied without damaging the cells of interest or reducing their fitness. Melanoma cell lines are cultured in Dulbecco’s modified Eagle’s medium (DMEM), which contains 4 mM Gln. We passaged 500,000 cells grown in complete medium into a Gln-deficient medium in 10-cm dishes; the only source of Gln was the 10% fetal bovine serum (FBS) used for supplementing the medium. Based on the results of amino acid analysis in calf serum, the level of Gln contributed by 10% FBS is about 27 M, which is not sufficient to meet the needs of proliferating cancer cells^13^. We frequently monitored the cells under a microscope, changing the medium as needed (e.g., to get rid of floating dead cells). We noticed that most of the cells died within a few days, and we observed no signs of cell proliferation.

We collected representative photographs of A375SM cells on dishes to document their status over time. Figure 1 shows the data from one of several experiments carried out with different numbers of cells for various lengths of time. We found a small number of cells remaining on a dish on day 8, and their number continued to decrease afterward. As a reference, if the cells were in complete medium, they grew to confluence by day 5. By day 27, very few cells remained in Gln-free medium; their morphologies were often abnormal, so whether they would persist and proliferate was uncertain (Fig. 1). However, upon further monitoring, we observed that some remaining cells gradually attempted to proliferate with different degrees of success. Eventually, some cells proliferated into colonies as demonstrated by the eight representative fields showing colonies on day 49 in Figure 1. Cells present in individual viewing fields exhibited significant differences in cell morphology and proliferation. Based on the number of cells present at different time points, surviving cells apparently were quiescent for about 4 weeks and then began exiting quiescence. The window for exiting quiescence may be wide based on heterogeneity in the depth of quiescence. We would not adequately score the cells that exited quiescence late (after other cells were ready to be passaged) or that did not exit it while staying alive. Of note was tremendous heterogeneity in the morphological features and speed of progression among individual cells as they went through survival in quiescence, with some of them gradually advancing to proliferation. This approach for modeling cells that persist in deep quiescence enabled us to establish long-term cultures of metabolically adaptable (MA) cells from A375SM cell line, which we thus far have passaged up to 10 times.

During these studies, we made an important observation: if we did not replace Gln-free medium for an extended period (more than 4–5 weeks), A375SM cells surviving in Gln-free medium appeared to stop exiting quiescence. In one experiment, we followed A375SM melanoma cells in Gln-free medium for 64 days. When we changed medium only once (no medium change for the last 50 days), cells got stuck in quiescence. The dish appeared blank upon crystal violet staining (Fig. 2 top left); however, we could see some single cells under microscope (Fig. 2 bottom left). We maintained another dish in Gln-free medium in parallel, the only difference being that we replaced old medium with fresh Gln-free medium 14 days before staining the dish with crystal violet (Fig. 2 top right). In this case one can see multiple cell colonies of different sizes.

We interpret that when the tiny amount of Gln (coming from FBS) is exhausted, the cells can still survive in quiescence. Once we provide them with fresh medium, small amount of Gln can support their growth into colonies. Perhaps the most important point is that an ability to proliferate in micromolar amount of Gln developed over time in quiescence under selection pressure, as it was lacking initially when cells were switched from complete medium to Gln-free medium (e.g., see Fig. 1).

Another strategy for extending quiescence in cell culture

Next, we increased the severity of the selection pressure/bottleneck to restrict cell proliferation by lowering the amount of Gln present in the culture medium to near 0 M by substituting regular FBS with dialyzed FBS. Dialyzed FBS also lacks other small molecular weight compounds normally present in FBS; we refer to these culture conditions as stringent Gln-free to distinguish them from regular Gln-free conditions using regular FBS. Upon microscopic monitoring of A375SM cells in stringent Gln-free medium, we noticed that quiescence was maintained for a markedly longer period than the one observed in regular Gln-free medium (compare cell images in Fig. 1 and Fig. 3a). Specifically, cells did not exit quiescence by week 7 in stringent Gln-free medium, whereas they exited quiescence after week 4 in regular Gln-free medium. One important question is whether lack of proliferation in stringent Gln-free medium results from irreversible severe damage to cells. To answer this question, after 28 days of incubation of A375SM cells in stringent Gln-free medium, we shifted the cells to complete medium with Gln and regular FBS. We found that upon removal of a severe bottleneck in this manner, cells exited quiescence and proliferated vigorously (Fig. 3b), indicating that they were arrested in deep quiescence but not severely damaged. Furthermore, as with the regular Gln-free conditions (Fig. 1), cells emerging from quiescence were heterogeneous in their morphology and rate of proliferation (Fig. 3b). It is likely that cells are heterogeneous regarding depth of quiescence and associated molecular characteristics.

Increased resistance to paclitaxel of rare melanoma cells surviving in deep quiescence

We refer to the rare cancer cells that survive prolonged, severe Gln deficiency by switching from cell proliferation to quiescence as MA for metabolic adaptability. Using a breast cancer model, we established that metabolic adaptability, an integral part of overall cellular adaptability, allows selection of highly adaptable cancer cells of the type that drive therapy resistance as well as recurrence and metastasis^5–7,14^. To assess a clinically relevant phenotype of A375SM-MA cells, we asked whether they are more resistant to paclitaxel than parental A375SM cells. Historically, paclitaxel has been used extensively in the treatment of melanoma, although it has gradually been replaced with immune checkpoint therapies and targeted therapies in recent years. A resistance to chemotherapeutic drugs such as paclitaxel in cell culture model of melanoma is a good indicator of therapeutic resistance of melanoma in the clinic. We treated A375SM-MA and parental A375SM cells with 5 nM or 10 nM paclitaxel for 6 days and then closely monitored them under a microscope to identify conditions that kill most cells with reasonable doses in a reasonable time. These conditions may also permit cell survival in quiescence, unlike treatments at high doses for a short time that are often used in cell culture studies. We found that A375SM-MA cells were considerably more resistant to paclitaxel than the parental A375SM cell line based on the number of cells present on dishes after 6 days of treatment (Fig. 4a). Next, to determine whether the cells surviving after 6 days of treatment could grow into colonies, we allowed paclitaxel-treated cells to recover in a medium without paclitaxel and grow into colonies for 9 days. We found that paclitaxel-treated A375SM-MA cells yielded significantly more colonies than did similarly treated parental cells, indicating a significantly higher chemotherapy resistance of A375SM-MA cells than parental cell line (Fig. 4b).

Survival of B16-BL6 mouse melanoma cells in quiescence under Gln deficiency

To determine whether the modeling approach described above for A375SM melanoma cells would work with another highly aggressive melanoma cell line, we chose the mouse melanoma cell line B16-BL6, which has been widely used in melanoma research. It is also suitable for evaluating therapies in syngeneic mouse models. During these studies, we recognized that B16-BL6 cells attach very strongly to the surface of cell culture dishes^15^, creating a technical difficulty in our approach. Specifically, observing cells on a cell culture dish for long periods is easier if only live cells stay attached to the dish than if both live and dead cells stay attached We learned that, possibly due to unusual cell membrane properties, even dead B16-BL6 cells or cell fragments stayed attached to the dish, thus creating a great deal of “background noise” as we observed rare live cells surviving in quiescence. Therefore, we had to make some adjustments of cell culture to monitor these cells. One adjustment was to passage the cells for getting rid of dead cells remaining attached to the dish.

Figure 5 shows two representative images of B16-BL6 cells on a dish after they had been in Gln-free medium for 12 days. We plated 500,000 B16-BL6 cells in Gln-free medium, with FBS providing only a low level of Gln (~27 M). Although most cells died in this medium, leaving behind only a small number surviving quiescence, monitoring of quiescence required an adjustment. Unlike the A375SM cell line, the B16-BL6 cell line adheres very strongly to cell culture dishes^15^. Even dead cells do not separate from a dish, making the monitoring of small numbers of these cells in quiescence very challenging. We noticed that cell growth was arrested, as the number of cells did not increase but rather declined over time. Also, an increase in the number of morphological abnormalities in these cells became more apparent. Control B16-BL6 cells placed in complete medium grew to confluence in 5 days. We passaged all cells from the dish to a new dish on day 7 with the intent of reducing background noise due to dead cells and collected photographs of cells on day 12. We found that most B16-BL6 cells die under the bottleneck created by Gln deficiency and that rare cells that survive in quiescence may not have yet succeeded in proliferation.

Some B16-BL6 cells may survive in low levels of Gln (from FBS in medium) for some time before they enter crisis and die. To address whether a low level of Gln could be extending cell survival we used stringent Gln-free medium prepared using dialyzed FBS. As anticipated, stringent Gln-free medium had a more severe effect on B16-BL6 cells than regular Gln-free medium (see representative images of cells obtained at several intervals up to day 48 in stringent Gln-free medium in Fig. 6). Most of the cells died sooner than they did in regular Gln-free medium, and rare cells that survived stayed arrested in quiescence (compare cells at different time points in Fig. 5 and Fig. 6a). Significantly, like the results obtained with A375SM cells (Fig. 3), B16-BL6 cells surviving in quiescence failed to exit it for up to 7 weeks. When we shifted the cells to complete medium for a week, cells exited quiescence and proliferated, indicating that the surviving cells were not irreversibly growth-arrested (Fig. 6b). Of note was a high degree of inter-colony and intra-colony heterogeneity of the cells.

To determine whether B16-BL6-MA cells that survive in deep quiescence and then proliferate in Gln-free medium may be capable of surviving other unrelated insults or challenges, we compared their ability to survive treatment with paclitaxel with that of parental B16-BL6 cells. Upon treatment with 20–80 nM paclitaxel in complete medium for 6 days, we found that B16-BL6-MA cells were considerably less affected than parental cells as demonstrated by comparing the numbers of cells in the different panels of Figure 7a. To compare paclitaxel’s effect in another manner (by comparing colony-forming ability/proliferation potential of cells that survived paclitaxel treatment), we shifted the paclitaxel-treated cells to drug-free medium for 9 days and then stained the dishes containing these cells with crystal violet (Fig. 7b). This provided an estimate of the proliferative ability of the cells remaining after 6 days of treatment with paclitaxel. The number of B16-BL6-MA cells surviving treatment with 80 nM paclitaxel, likely in quiescence and then growing into colonies in drug-free medium, was dramatically higher than that of parental B16-BL6 cells. Taken together, these results demonstrated that B16-BL6-MA cells are more resistant to paclitaxel than parental cell line because of their ability to survive in deep quiescence.

One noteworthy point is that after culturing MA cells in Gln-free medium, they continue to grow significantly slower than the parental cell line does in complete medium. Therefore, besides demonstrating the ability of adaptable melanoma cells to survive in quiescence, our model could also serve as one of slow-cycling cancer cells. That these two traits of cancer cell survival in quiescence and their slow-cycling nature go together appears to be reasonable. One way to think about it is that there may be a continuum of cell states with increasing degree of cell proliferation between quiescence and proliferation.

Cancer evolution can be viewed as driven by a stem-like cell state corrupted by genetic and epigenetic alterations^16–19^. The tumor microenvironment is also an integral part of cancer evolution^20^. In fact, cancer cells and their microenvironments have a reciprocal relationship. Depending on the context of the disease, the microenvironment can either inhibit or promote cancer; cancer can influence or, specifically, corrupt the microenvironment. A reasonable notion is that the microenvironment is on the winning side keeping cancer in check at the initial residual disease stage. However, the balance gradually shifts in favor of cancer close to the time of recurrence. Our goal in this study was to model the cancer cell state that enables cell survival and further evolution of melanoma under all restrictive environments imposed by the body to enforce a code of multicellularity that favors the organism over individual cells^11^. Such a cancer cell state would be responsible for recurrence and metastasis.

Although the fact that a cell culture model can be used as a model of cancer cell dormancy may appear counterintuitive, some serendipitous observations, initially while investigating aggressive TNBC, have led us in this direction. Preclinical models, including cell line models, are notoriously poor at predicting cancer’s response to treatment in the clinic. One reason for this is that various bottlenecks in the body eliminate proliferating cancer cells, allowing for selection and further evolution of rare cells that are good at opportunistically switching between quiescence and proliferation. In contrast, because cell culture conditions do not impose such a bottleneck, most cells that would not have survived in the body continue proliferating in cell culture. Our study demonstrated that prolonged, severe metabolic challenge can serve as a suitable bottleneck for selecting rare cancer cells like the ones responsible for recurrence. Perhaps equally important, the bottleneck can be used for eliminating the bulk of proliferating cells in cell culture that are not adaptable and thus are likely to be unimportant in the disease context clinically.

In our phenotype-based modeling of adaptable cancer cells, which involves persistence under a severe selection pressure, we expected a high degree of molecular heterogeneity among cancer cells. Each cell that survives in quiescence may have a unique evolutionary history and unique trajectory for exit from quiescence and unlimited proliferation. This is supported by morphological observations in our study indicating a lack of uniformity of quiescence among surviving cells, as cells may go through multiple steps—some futile, some successful—for survival in quiescence and exit from it. It is also supported by molecular characterization of MA cells in similar studies of highly aggressive triple-negative inflammatory breast cancer^6,7,14,21^. Our results are consistent with a model in which a shortage of Gln creates a bottleneck for cell survival and proliferation, thus forcing selection of stem-like cancer cells based on their ability to survive in deep quiescence (Fig. 8). Similarly, a shortage of nutrients often serves as a niche for various types of stem cells in adult tissues. Notably, others have shown that reduced dependence on exogenous Gln for cell survival and proliferation is a general feature of self-renewing pluripotent stem cells^22^.

Cell culture is frequently used for modeling the cancer cells important to cancer progression and therapy resistance. Therefore, comparing our approach to other approaches is reasonable. The differences are based primarily on our focus on cancer cells that persist in the body, often in quiescence, and eventually cause recurrence^10,23,24^. Examples of other approaches include 1) enrichment of cells capable of metastasizing to specific organs by injecting cells into nude mice followed by culturing cells from metastatic sites, 2) selection of therapy-resistant cells based on their ability to survive prolonged treatment with a particular agent, 3) selection of stem-like cancer cells based on their ability to grow as spheres under low-attachment culture conditions, and 4) selection of invasive/metastatic cancer cells based on their ability to grow in soft agar, reflecting an altered extracellular matrix and capacity to invade. All these approaches are useful to a degree, but they do not optimally model a switch to quiescence, which is a major goal of our study. In this regard, some laboratories investigate quiescence of cancer cell lines primarily by withholding growth factors or serum from the culture medium for a relatively short period (few days). In our opinion, which is supported by various studies including our own, this approach models only shallow quiescence. It is well known that the amount of time a cell stays in quiescence correlates with the depth of quiescence.

As Poste and colleagues concluded, “the level of expression of the invasive phenotype in vivo may be determined by the severity of the selection procedure in vitro”^25^. Although this conclusion was based on selection procedures different from ours, expanding the choice of selection procedures to those involving severe and prolonged metabolic challenges such as ours is reasonable. A fundamental difference between other approaches designed to model a specific phenotype (e.g., cell invasion) and our approach is that rare cells that can survive in abnormal quiescence can generate a variety of cancer cell phenotypes that enable them to persist in foreign environments, eventually causing recurrence and/or metastasis.

Our approach of modeling residual disease on a dish in Gln-free medium is well-suited for aggressive cancer cell lines representing undifferentiated tumors since they are highly addicted to Gln for cell proliferation. In contrast, cell lines representing differentiated tumors (e.g., estrogen receptor-positive breast cancer cell line MCF7) can easily survive a lack of Gln without a long arrest in quiescence^5^. This is consistent with the non-essential nature of Gln, which implies that most normal cells in the body do not need a high amount of exogenous Gln for their functionality.

An important consideration in cancer dormancy is the nature of cancer cell quiescence. First, cells that have been in quiescence for a long period are known to normally take longer to exit from it than the ones that have been quiescent for a short period. Long and deep quiescence is accompanied by specific molecular features associated with condensed and super condensed chromatin as well as those responsible for maintenance of various subcellular structures^26^. Although we can rely on knowledge of quiescence in various normal contexts in nature for understanding the molecular mechanisms of quiescence in cancer cells, many of the mechanisms governing quiescence are corrupted in cancer cells. To deal with this complexity, our approach is to model a deep abnormal quiescence phenotype, which would render quiescence more like the disease prior to recurrence. At this stage of model generation, rather than relying on specific biomarkers of quiescence for tracking quiescent cancer cells, as good biomarkers that can be universally analyzed in tremendously heterogeneous and evolving cancer cell populations are lacking, we rely on lengthy monitoring of cells in culture to determine their depth of quiescence. This straightforward approach works well in our cell culture model for selecting and evaluating cells in deep quiescence.

One clinically relevant question may be whether the phenotype modeled as described herein is relevant to resistance to current therapies, including immune checkpoint therapies. Although we have yet to evaluate the effects of various modes of therapy in our model besides paclitaxel, authors have reported strong evidence that the ability of cancer cells to survive in deep quiescence is a major hurdle for all therapies, including immune checkpoint therapies^27^. Therefore, we strongly believe that progress in overcoming intrinsic treatment resistance by targeting abnormal deep quiescence in cancer cells will have a significant impact to improving therapeutic strategies for resistant cancers such as melanoma.

Methods

Cell lines and drugs

Two highly metastatic melanoma cell lines were used in this study. A375SM is a human melanoma cell line derived from the A375 cell line established in culture from a lymph node metastasis that was intravenously injected into nude mice, and the cells that metastasized to the lung were cultured for establishing A375SM cell line^9^. B16-BL6 is a mouse melanoma cell line that underwent 10 rounds of selection for lung colonization in mice followed by in vitro selection for enhanced tissue invasion^25^. We purchased both the cell lines from UT MD Anderson Cancer Center Cytogenetics and Cell Authentication Core Facility. Both these cell lines were established by other investigators many years ago. Our Core facility now maintains cell lines.

Paclitaxel, which was purchased from Sigma-Aldrich (St. Louis, MO, USA), was dissolved in dimethyl sulfoxide (DMSO). Equal volume of DMSO without paclitaxel was added to the control dishes. The DMSO volume was up to 0.04% of the culture medium volume.

Selection of melanoma cells surviving in quiescence in a bottleneck

A prolonged lack of Gln in culture medium was chosen as a bottleneck because it is strong but does not kill all melanoma cells. The rare cells (~0.01% of the population) that survived the bottleneck did so by surviving in deep quiescence for several weeks. Our monitoring of such resistant cells involved observing them under a microscope for long periods (several weeks). Because universal markers for highly abnormal stem-like cancer cells persisting in deep quiescence are lacking, we relied on microscopic monitoring of cellular quiescence. Similarly, to determine whether cancer cells can exit quiescence (overcome a bottleneck), we observed the cultures for long periods to determine whether they had evidence of cell proliferation. This function-based assay of a quiescent cell phenotype is optimal for a deeply quiescent abnormal cancer cell phenotype.

The bottleneck was applied to melanoma cells growing in cell culture by leaving Gln out of the cell culture medium. Because the FBS used as a supplement in the medium had only a small amount of Gln (27 M, compared with 4 mM in regular DMEM), this source of Gln was eliminated from the cell culture medium by using dialyzed FBS instead of regular FBS. Results obtained using both media are presented herein.

Assay of relative resistance of melanoma cells to paclitaxel

The assay of relative resistance of MA cells versus parental cells was optimized to assess the effect on cell survival of reasonably low doses of paclitaxel that would permit the survival of highly resistant subpopulations of melanoma cells in quiescence for several days by first evaluating a broad range of drug concentrations for a long period (1–2 weeks). A375SM and A375SM-MA cells were treated in parallel with vehicle-treated control cells for 6–7 days with predetermined concentrations of paclitaxel (5 or 10 nM) expected to kill about 99% of proliferating parental cells. Drug-containing medium was then removed from cell culture dishes, and the cells twice with phosphate-buffered saline solution and incubated in the medium without paclitaxel for 7–14 days until colonies visible to the naked eye appeared. Cell colonies were stained with crystal violet^28^ and counted for quantitation of data.

The relative resistance to paclitaxel of B16-BL6-MA and parental B16-BL6 cells was evaluated in a manner like that for A375SM-MA and A375SM cells. However, because mouse melanoma cell line B16-BL6 is more resistant to paclitaxel than human cell line A375SM, paclitaxel in the 20–80 nM range was used.