Lat1 (SLC7A5) is an amino acid transporter often required for tumor cell import of essential amino acids (AA) including Methionine (Met). Met is the obligate precursor of S‐adenosylmethionine (SAM), the methyl donor utilized by all methyltransferases including the polycomb repressor complex (PRC2)‐specific EZH2. Cell populations sorted for surface Lat1 exhibit activated EZH2, enrichment for Met‐cycle intermediates, and aggressive tumor growth in mice. In agreement, EZH2 and Lat1 expression are co‐regulated in models of cancer cell differentiation and co‐expression is observed at the invasive front of human lung tumors. EZH2 knockdown or small‐molecule inhibition leads to de‐repression of RXRα resulting in reduced Lat1 expression. Our results describe a Lat1‐EZH2 positive feedback loop illustrated by AA depletion or Lat1 knockdown resulting in SAM reduction and concomitant reduction in EZH2 activity. shRNA‐mediated knockdown of Lat1 results in tumor growth inhibition and points to Lat1 as a potential therapeutic target.
A metabolic–epigenetic feedback loop between the methionine transporter Lat1 and the histone methyltransferase EZH2 dictates differentiation state in a malignant context.
Cells sorted for high CD98/Lat1 expression contain elevated methionine cycle metabolites and more active EZH2 and are more aggressive in lung cancer models.
Amino acid restriction, methionine dropout, or Lat1 shRNA impairs EZH2 activity due to reduction in cellular S‐adenosylmethionine.
Lat1 and EZH2 expressions correlate with a less‐differentiated state as illustrated by spheroid models of cancer cell differentiation, retinoic acid‐mediated transcription, and immunohistochemical analysis of human lung cancer.
EZH2 activity and PRC2 activity derepress Lat1 expression through direct promoter binding and subsequent transcriptional repression of RXRα.
It is widely recognized that malignant cells exhibit unique metabolic profiles consistent with elevated anabolic demands as well as adaptation to elevated cellular stress and changes in the microenvironment. Some of these metabolic traits have been exploited by oncologists for both diagnostic and therapeutic intervention. For instance, the high glucose avidity of tumors provides the basis of 2‐deoxy‐2‐(18F)‐D‐glucose‐based positron emission tomography imaging. Another example exists in the elevated rate of nucleotide biosynthesis rendering malignant cells vulnerable to drugs like pemetrexed and 5‐fluorouracil, which target enzymes in the purine and pyrimidine synthesis pathways.
Over the past decade, cancer sequencing efforts have uncovered a number of genetic alterations in metabolic enzymes. Loss‐of‐function mutations in succinate dehydrogenase (SDHA) and fumarate hydratase (FH) result in accumulation of their substrates succinate and fumarate, respectively. In addition, neomorphic gain of function mutations in IDH1 and IDH2 leads to accumulation of 2‐hydroxyglutarate (2‐HG) (Dang et al, 2009; Gross et al, 2010; Xiao et al, 2012). In these circumstances, the accumulating metabolite acts as a competitive inhibitor of α‐ketoglutarate (α‐KG)‐dependent enzymes including dioxygenases that affect histone and DNA methylation. Metabolite‐mediated inactivation of histone or DNA demethylases reprograms the cancer epigenome, which underscores the fundamental link between cellular metabolism and epigenetics (Sasaki et al, 2012; Xiao et al, 2012).
While a number of studies have focused on the characterization of the mechanisms modulating the activity of epigenetic demethylases, less is known about how metabolism regulates the epigenetic enzymes involved in methyltransferase reactions in cancer cells. All cellular methyltransferases depend on the methyl donor S‐adenosylmethionine (SAM) for activity. The catalytic subunits of the methionine adenosyltransferase (MAT1A & 2A) complexes form SAM by combining a molecule of ATP and Met (Lu & Mato, 2012). The amino acid transporter SLC7A5 (herein referred to as Lat1) has been previously characterized by its substrate selectivity for the branched chain amino acids and bulky amino acids including Met (Mastroberardino et al, 1998). This transporter functions at the cell surface in covalent complex with its obligate chaperone SLC3A2 (herein referred to as CD98). In a number of tumor types, Lat1 expression predicts poorer overall survival and a less‐differentiated state (Ring et al, 2006; Kaira et al, 2008; Yanagisawa et al, 2012). Further, previous reports describe Lat1 upregulation in response to activation of Myc (Gao et al, 2009).
Here, we describe a novel role for Lat1 as an indirect regulator of a cell's epigenetic state by restricting the availability of Met, and thereby reducing the available pool of SAM. Reduction in the cellular SAM pool results in a loss of several histone methyl marks including the EZH2‐specific histone H3, lysine 27 trimethylation (3meH3K27), known to play a fundamental role in regulating the expression of genes involved in differentiation (Lee et al, 2006). We also describe a feedback loop where EZH2 activity regulates the expression of Lat1 through PRC2‐dependent repression of retinoid receptor α (RXRα). Our data suggest that interfering with the activity of amino acid transport may impact the activity of epigenetic regulators and in turn affect transcriptional programs under their control.
Cell populations enriched for Lat1/CD98 surface complexes exhibit higher Met‐cycle metabolite levels
We sought to metabolically profile heterogeneous lung cancer cell subpopulations as dictated by cell surface expression of the Lat1/CD98 transporter. We took advantage of the well‐characterized antibodies specific for CD98, and fluorescent‐activated cell sorting was performed on NCI‐H1299 and NCI‐H520 cancer cell lines to enrich for Lat1 complexes (Fig 1A and B). NCI‐H1299 cells enriched for CD98/Lat1 display a more aggressive and invasive phenotype when grown on top of matrigel compared to CD98lo and also form more aggressive tumors in immune‐compromised mice (Fig 1C and D). While this is in agreement with previous reports where the CD98/Lat1 amino acid transporter was linked to stem‐like properties of cancer cells (Martens‐de Kemp et al, 2013), it should be noted that both CD98hi and CD98lo cells were able to form tumors with the same penetrance (9/10 mice). In light of these data, we interpret the enrichment or expression of CD98/Lat1 as a measure on a spectrum of cellular differentiation as opposed to a stem‐cell state.
Using a targeted metabolomics analysis approach, we phenotyped CD98hi/lo NCI‐H1299 and NCI‐H520 cells and found significant elevation of several Met‐cycle intermediates including SAM, S‐adenosylhomocysteine (SAH), methylthioadenosine, adenosine, and cystathionine in CD98hi cells (Fig 1E). In this analysis, Met itself does not appear to change; however, our sample collection process requires that cells are kept in serum‐free media prior to metabolite extraction, which dilutes the cell autonomous read‐out for essential amino acids. To overcome this technical limitation, we took advantage of ion exchange chromatography to specifically assay intracellular Met concentrations from CD98 sorted cells. Using this methodology, we find that CD98hi NCI‐H1299 cells contain higher concentrations of free Met, a known Lat1 substrate (Fig 1F).
CD98hi NCI‐H1299 and NCI‐H520 cells exhibit an enrichment of 3meH3K27, the EZH2‐specific methyl mark associated with stem‐cell‐like, less‐differentiated features in cancer (Fig 1A and B; Iliopoulos et al, 2010). In addition, sorted CD98hi NCI‐H520 cells propagated in culture after sorting display persistent enrichment of both EZH2 and 3meH3K27 (Fig 1F). While CD98hi sorted NCI‐H1299 cells were not enriched for total EZH2, we did observe enrichment of phosphorylated EZH2 at T487 (Fig 1A). This phosphorylated form of EZH2 is indicative of higher activity as previously observed using a recombinant phospho‐mimetic EZH2 mutant at this site (Fig 1A; Kaneko et al, 2010). In addition to a lack of detectable EZH2 pT487, CD98lo NCI‐H1299 cells express lower levels of MAT2A, the catalytic subunit of the MATII complex which may contribute to the observed reduction in histone methylation, Met and SAM in these cells (Fig 1A).
Given the enrichment for markers indicative of a less‐differentiated state in CD98hi cells, we analyzed indicators of epithelial to mesenchymal transition (EMT), a cellular process previously shown to correlate with cancer cell stemness (Supplementary Fig S1A). While we did not see changes in the expression of E‐cadherin, Slug, or Snail between CD98lo and CD98hi NCI‐H1299 cells (Supplementary Fig S1A), we did find accumulation of actin stress fibers in the CD98hi fraction (Supplementary Fig S1B). The stress fiber phenotype is indicative of cytoskeletal remodeling that could account for the more aggressive growth of CD98hi cells in vivo.
Lat1 and EZH2 expression are reflective of differentiation state
The metabolic, epigenetic, and in vitro/in vivo growth phenotypes observed in cells sorted on the basis of cell surface CD98/Lat1 suggest that expression of this complex may be linked to differentiation state. Moreover, expressions of Lat1 and EZH2 are part of a previously identified gene signature that describes an undifferentiated tumor state (Rhodes et al, 2004). We sought to evaluate whether any correlation exists between the expression of these genes in in vitro models where differentiation can be induced (i.e. spheroid growth & pro‐differentiation culture media) or suppressed (i.e. retinoic acid restriction & SV40 large‐T antigen expression).
Cancer cell lines can be differentiated through adaptation to 3‐dimensional (3D) culture conditions as spheroids (Sutherland, 1988). We find that multiple non‐small cell lung cancer (NSCLC) cell lines will adapt to spheroid growth when grown on non‐adherent plates in the presence of 2% extracellular matrix (ECM). We compared expression in NCI‐H1299 cells grown on plastic and as spheroids and found various changes reflective of a more differentiated state (Fig 2A and Supplementary Fig S1C). For instance, the Integrin/FAK pathway was largely downregulated upon spheroid growth and markers of epithelial to mesenchymal transition (EMT) were consistent with epithelial characteristics (Supplementary Fig S1C). Loss of Lat1, CD98, and EZH2 was also observed upon adaptation to spheroid culture (Fig 2A). We next assayed the expression level of the same proteins in the primary NSCLC cell line TUM622 when grown in adherent and spheroid conditions (Damelin et al, 2011). Again, we observed a loss of Integrin/FAK pathway signaling, a shift to a more epithelial phenotype and concomitant loss of EZH2, Lat1, and CD98 (Fig 2B and Supplementary Fig S1D).
TUM622 cells grow in defined media, which allowed us to evaluate the effect of retinoic acid (RA), a well‐characterized pro‐differentiation factor, on the expression of Lat1 and EZH2. We found that upon removal of RA from culture media, Lat1 and CD98 expressions, but not EZH2, were elevated. In the spheroid cultures, Lat1 and CD98 levels were rescued to levels observed in adherently growing cells (Fig 2B). Further, when TUM622 cells are grown in media purposed for propagation of differentiated airway epithelial cells, a reduction in both Lat1 and EZH2 as well as migration of CD98 to a lower molecular weight species is observed over time (Fig 2C). In this setting, expressions of Lat1 and EZH2 are inversely correlated with Involucrin, a known marker of terminal differentiation in squamous epithelial cells (Lee et al, 1996).
To rule out differences that may occur due purely to changes in cell cycle, we evaluated the changes in Lat1 and EZH2 expression upon nocodazole block and release in NCI‐H1299 cells. No difference in Lat1 levels was observed during 24‐h cell cycle progression. On the other hand, EZH2, Cyclin B1, and Cyclin D1 levels did vary over time consistent with previous reports (Fig 2D; Wei et al, 2011). Further, growth of TUM622 cells in the absence of RA had no significant effect on cell proliferation over time as assayed by viable cell counts (Supplementary Fig S1E).
In an effort to develop an independent model to explore the relationship between CD98, Lat1, and EZH2 in the transition from the differentiated to undifferentiated state, we immortalized normal human bronchioalveolar cells (NHBE) with lentiviral‐delivered SV40 large‐T antigen (Fig 2E). In this experimental setting, we again see predictable response in the Integrin/FAK pathway and EMT markers as well as robust elevation of Lat1, CD98, and EZH2 expression (Supplementary Fig S1F and Fig 2E). Of particular note, we observed a correlation of MAT2A expression with Lat1 and EZH2 reminiscent of CD98hi sorted cells suggesting that the Met‐cycle is upregulated in less‐differentiated states (Fig 2B and E).
Lat1 expression is regulated by EZH2 through de‐repression of RXRα
Given the correlative link between EZH2 and Lat1 as well as the clear effect of RA on CD98/Lat1 expression described above, we sought to determine the impact of EZH2 loss on Lat1 expression and Met levels. To that end, we constitutively knocked down EZH2 in NCI‐H1299 cells. Western blot analysis confirmed the loss of EZH2 and 3meH3K27, and downregulation of CD98/Lat1 (Fig 3A). The observed reduction in Lat1 expression reflects a functional loss as cellular levels of Met are reduced in cells with EZH2 knockdown (Fig 3B). Furthermore, treatment of NCI‐H1299 cells with the EZH2 inhibitor GSK126 resulted in a reduction in Lat1 expression (Fig 3C). The data provide pharmacological confirmation of the connection between EZH2 activity and Lat1 expression.
In order to determine whether Lat1 expression is directly regulated by EZH2 activity, we performed chromatin immunoprecipitation (ChIP) with antibodies against EZH2, SUZ12, 3meH3K27, and RNA Polymerase II (as a positive control of active transcription). We did not observe enrichment of PRC2 complex members or 3meH3K27 at the Lat1 or CD98 promoter indicating that expression of Lat1 and CD98 are regulated by PRC2 via an indirect mechanism (Fig 3D). To ensure our ChIP experiments were properly assaying PRC2 promoter occupancy, we used CCND2 as a positive control PRC2 target gene (Supplementary Fig S2A).
The link between EZH2 and RA‐induced differentiation has been previously reported (Epping et al, 2005), and the effect of RA on CD98/Lat1 expression appeared to be downstream of EZH2 (Fig 2B). Therefore, we also investigated whether PRC2 directly regulated retinoic acid receptor and retinoid X receptor genes. ChIP experiments revealed significant enrichment of EZH2, SUZ12, and 3meH3K27 at the promoters of RARβ and RXRα, but not RARα and RXRβ in H1299 cells (Fig 3E). In addition, shRNA‐mediated knockdown of EZH2 resulted in selective upregulation of RARβ and RXRα, confirming that these genes are directly repressed by the PRC2 complex in these cells (Fig 3G and Supplementary Fig S2B).
Based on these results, we postulated that the EZH2‐mediated effect on Lat1 regulation was through direct regulation of either RARβ or RXRα. To test this hypothesis, we used siRNA to deplete either RARβ or RXRα in cells where constitutive knockdown of EZH2 resulted in loss of Lat1 expression. We found that inhibition of RXRα, but not RARβ, rescued Lat1 mRNA and protein to the levels observed in non‐targeting control cells (Fig 3F and G and Supplementary Fig S2B). Further, RXRα siRNA alone was sufficient to elevate Lat1 expression at both the mRNA and protein level reminiscent of the RA dropout experiment described above (Fig 3F and G). To confirm that this was, in fact, PRC2 complex dependent and not EZH2 autonomous activity, we repeated the RXRα siRNA experiment during SUZ12 shRNA knockdown and observed similar effects on Lat1 mRNA expression (Supplementary Fig S2C). Finally, we confirm that RXRα binds the Lat1, CD98, and MAT2A promoter, while RARβ does not contribute any promoter binding even to the positive control ABCA1 gene in these cells (Supplementary Fig S2D and E). Taken together, these data revealed a role for RXRα as a negative regulator of Lat1 and suggest that the PRC2 complex potentiates Lat1 expression by directly repressing RXRα gene expression.
Lat1 and EZH2 co‐segregate in NSCLC clinical samples as a function of tumor stage
To understand the clinical relevance of the preclinical observations previously described, we examined EZH2 and Lat1 expression across gene expression data sets obtained from samples of patients with lung cancer. Using both proprietary (GeneLogic) and external (Oncomine) mRNA expression databases, we found that both EZH2 and Lat1 are upregulated in NSCLC tumor samples (both adenocarcinoma and squamous cell carcinoma subtypes), while the related gene homologues EZH1 and SLC7A7 (i.e. y+Lat1) are down regulated (Fig 4A and B, Supplementary Fig S3A and B). We developed IHC methodology to assay for EZH2 and Lat1 expression in tissue microarrays of samples of patients with NSCLC of various stages. Statistical analysis of staining patterns shows a significant overlap of Lat1 with EZH2 expression (Fisher's exact test P = 1.98E−04), and expression of each also significantly correlates with increasing stage of tumor (Fisher's exact test P = 4.45E−08 and 1.78E−07, respectively; representative images are shown in Fig 4B). Co‐staining individual samples for EZH2 and Lat1 in lung squamous cell carcinoma reveal that these proteins are co‐expressed in the same cell populations. Specifically, we observed robust staining for both EZH2 and Lat1 within the least differentiated, most rapidly proliferating, tumor cells adjacent to the stroma (Fig 4C).
Loss of Lat1 by shRNA knockdown limits Met‐cycle metabolites and reduction in H3K27 methylation
Based on the observed interplay between metabolic and epigenetic effectors, we reasoned that epigenetic state could be modulated simply by restricting the cellular availability of amino acids (AA). Given the ability of Lat1 to transport Met substrate, we sought to evaluate its functional role in terms of cellular Met availability and growth. We find that shRNA‐mediated knockdown of Lat1 in NCI‐H1299 lung cancer cells results in a significant reduction in Met (Fig 5A) as well as total SAM concentrations (Fig 5B). Based on these data, we next interrogated the effect of Lat1 knockdown on the EZH2 substrate H3K27 and observed a reduction in H3K27 di‐ and tri‐methylation (Fig 5C). Importantly, we find a number of histone methyl marks are reduced upon knockdown of Lat1 including substrates for MLL1 (H3K4) and MMSET (H4K20), suggesting that SAM substrate channeling from Lat1‐specific Met entry is not limited to EZH2 activity (Supplementary Fig S4A). In order to substantiate a model of Met restriction on the cellular SAM pool and epigenetic state, we used shRNA to knockdown MAT2A and interfere directly with the Met‐cycle. Due to the critical and non‐redundant role of MAT2A in multiple cellular processes, it was essential to employ a doxycycline‐inducible shRNA system to complete these experiments. We find that depletion of MAT2A phenocopies loss of Lat1, leading to a comparable reduction in 3meH3K27 (Fig 5D). Based on the shRNA experiments described above, it follows that NCI‐H1299 cells depleted of AAs will respond with reduced 3meH3K27 (Fig 5E). Using an unrelated cell line (HEK‐293), we find that 3meH3K27 reaches a minimal level at 4 h post‐depletion thereby showing that amino acid dependency on 3meH3K27 is not limited to the lung cancer models used in these studies (Fig 5F). Moreover, restriction of Met alone is sufficient to reduce 3meH3K27, whereas restriction of a control amino acid (cysteine) has no effect on epigenetic state (Fig 5G).
Lat1 depletion interferes with NCI‐H1299 tumorigenicity
Finally, we investigated the contribution of Lat1 to cancer cell growth using shRNA knockdown in NCI‐H1299 cells. Although knockdown of Lat1 did not affect cell growth on plastic, the same hairpins impaired NCI‐H1299 cell growth when plated on top of extracellular matrix reminiscent of the CD98hi sorted cells (Fig 5G). We also find that shRNA‐mediated loss of Lat1 results in impaired invasion through matrigel‐coated Boyden chambers, suggesting that at the invasive tumor front, Lat1 may facilitate the expansion of tumor boundaries into the surrounding tissue (Supplementary Fig S4B). Ultimately, we sought to determine whether loss of Lat1 would result in significant tumor growth inhibition in vivo. We observed NCI‐H1299 tumor growth inhibition consequent to Lat1 shRNA‐mediated knockdown (Fig 5I). Likewise, induction of MAT2A shRNA in the same tumor model also manifested robust tumor growth inhibition strengthening the model that Lat1 and MAT2A operate within the same metabolic/epigenetic axis (Supplementary Fig S4C).
Malignant cells rewire their transcriptional, signaling, and metabolic networks to suit the biosynthetic demands of unchecked proliferation. It is apparent that these networks are interconnected and highly responsive to changes in one another. For instance, mutations in metabolic genes such as IDH1, IDH2, SDH, and FH lead to alterations in metabolites that interfere with the activity of dioxygenases, including histone demethylases, by virtue of competition with α‐KG. As a consequence, the broad epigenetic imbalance results in transcriptional and signaling deregulation (Xiao et al, 2012). Fluctuations in a number of metabolites have been shown to affect a wide range of epigenetic regulators. For example, in addition to its substrate function within the Kreb's cycle and as an anabolic building block, Acetyl‐CoA levels also directly impact histone acetylation (reviewed in Kaelin & McKnight, 2013). In yeast, Acetyl‐CoA concentrations in the cell dictate the activity of histone acetyltransferases such as GCN5 leading to broad dynamic changes in transcriptionally active chromatin (Cai et al, 2011). Additionally, in mammalian cells, histone acetylation requires the conversion of citrate into acetyl‐CoA by ATP‐citrate lyase (Wellen et al, 2009). Further, removal of histone acetylation marks can be mediated by members of the sirtuin family, which are dependent on another metabolite, nicotinamide adenine dinucleotide (NAD+) as a cofactor for the deacetylation reaction (Kaelin & McKnight, 2013).
Here, we describe a previously unknown and reciprocal regulation between extracellular Met uptake and epigenetic state (modeled in Fig 6). Our data suggest that Lat1 is critical for transport of the essential amino acid Met and maintenance of the steady state cellular SAM concentration. Reduction in SAM levels correlates with loss of Lat1 as well as the histone methylation mark catalyzed by EZH2, 3meH3K27. In accordance with these observations, expressions of EZH2 and Lat1 are tightly correlated in cell culture models of differentiation. Furthermore, expression of both proteins is high and coincident in clinical samples of lung cancer compared to adjacent normal tissue and positively correlates with tumor stage. Our study also shows that an enrichment of the CD98/Lat1 complex specifies a cancer cell population inherently more aggressive in vitro, reflected in their organization of actin stress fibers and growth on top of matrigel and in vivo reflected by their accelerated tumor growth kinetics. The CD98/Lat1hi cell characteristics are correlated with high EZH2 activity. In agreement, CD98/Lat1hi cells have elevated Met‐cycle intermediates indicative of their requirement for SAM in modulating EZH2 epigenetic effects.
In a reciprocal feedback relationship, we also identify Lat1 as a downstream biomarker of EZH2 activity. Our first indication of this relationship came from observations in spheroid cell cultures, where Lat1 and EZH2 levels drop but only Lat1 decreased in a RA‐dependent manner. Loss of PRC2 activity by either shRNA‐ or small‐molecule‐mediated inhibition also reduced Lat1 levels in cells, and this effect is abrogated in the absence of RXRα. We postulate that RA binds RXRα resulting in activation and DNA binding of this nuclear hormone receptor. That RXRα can influence the induction, and in this case, suppression of target genes upon activation by RA is likely through distinct binding partners at genetic loci associated with differentiation or progenitor phenotypes, respectively.
These data can be explained mechanistically by the observation that EZH2, SUZ12, and 3meH3K27 (markers of transcriptional repression) but not RNA Polymerase II (indicative of active transcription) were found at the RXRα promoter. Of particular note, we recognize a disconnect where EZH2 knockdown results in a modest reduction in Lat1 mRNA but a robust reduction in Lat1 protein. This suggests that EZH2 plays a role in Lat1/CD98 protein stability perhaps through RXRα‐independent regulation of degradation pathways. Development of small‐molecule candidates targeting EZH2 and their utility in the clinic will ultimately determine whether Lat1 expression can act as a surrogate biomarker for inhibitor response.
Finally, we present data that underscore the critical role of Lat1 in supporting tumor growth. The anti‐tumor effect mediated by loss of Lat1 was only evident after day 7 when most tumors reached an average size of 100 mm3. One potential explanation for this observation is that a threshold tumor size may exist at which malignant cells become more dependent on Met uptake. We hypothesize that as tumors grow and increase their Met demand for protein and SAM synthesis, they become more dependent on Lat1. Interfering with this transporter compromises the ability of tumor cells to replenish the Met pool resulting in metabolic crises. It should be noted that Km of EZH2 for SAM has been reported in vitro at approximately 1.2 μM, while the Ki for SAH is approximately 7.5 μM (Richon et al, 2011). Given the likely higher affinity for SAM, we would predict higher EZH2 activity in a SAM‐enriched cell. The ratio of SAM:SAH also plays a role in allosteric regulation of EZH2 and other methyltransferases in a cellular context and remains a caveat when interpreting our results.
Corroborating these data is the finding that MAT2A knockdown also results in robust tumor growth inhibition in the same model. Interestingly, while MAT2A expression does not appear to be affected by loss of EZH2, its promoter is bound by RXRα suggesting some EZH2‐independent but RXRα‐dependent regulation of this metabolic enzyme. In order to exploit SAM metabolism as a therapeutic strategy in oncology, it will be of utmost importance in future studies to fully evaluate the contribution of various methionine endpoints (i.e. translation, epigenetics, SAM‐dependent biosynthesis, etc.).
Materials and Methods
Cell lines, spheroid, and 3D cell culture assays
HEK293, NCI‐H1299, and NCI‐H520 cell lines were purchased from ATCC and grown in cell culture as per ATCC recommendations. Derivation of TUM622 cells was previously described and grown as per author's recommendation. Primary Normal Human Bronchioalveolar Epithelial (NHBE) cells were purchased from Lonza and subcultured as per the manufacturer's recommendation.
Experiments involving NCI‐H1299 or TUM622 cells grown as pulmospheres were as follows. 10,000 cells were plated in full media containing 2% extracellular matrix (Matrigel) on a low binding 6‐well plate. Spheroid morphology and growth was monitored over the course of 10–15 days and photographed as explained below.
Experiments involving NCI‐H1299 cells grown on top of matrigel were as follows. 300 μl of ice‐cold 80% matrigel in full media was plated into a 24‐well cell culture dish and allowed to harden for 2 h at 37°C. NCI‐H1299 cells were treated with siRNA or sorted for CD98 as described, and 100,000 cells were seeded in 1 ml of full media onto the matrigel plate. Cell morphology and growth was monitored over the course of 10–15 days and photographed as explained below.
Experiments involving cellular invasion assays were performed using the Cultrex BME Cell Invasion assay as per the manufacturer's suggested protocol (Trevigen).
siRNA mediated knockdown
Transient knockdown experiments were carried out as per the manufacturer's recommendation. Briefly, Silencer Select siRNA oligonucleotides (Life Technologies) targeting SLC7A5 or RXRA were transfected with Lipofectamine™ RNAiMAX. NCI‐H1299 cells were seeded 100,000 cells/well in a 6‐well plate 24 h prior to transfection. Oligonucleotide and transfection reagent were complexed according to the manufacturer's instructions and delivered to cells at a final concentration of 5 nM. siRNA oligonucleotide sequences used in this study are found in Supplementary Table S1.
Fluorescent‐activated cell sorting
Adherent cells were trypsinized and washed with PBS prior to incubation with mouse monoclonal anit‐CD98 (Supplementary Table S2) antibody for 1 h at 4°C. Excess primary antibody was washed off in a 1% FBS/PBS solution three times followed by further incubation with PE‐tagged secondary antibody for a further 30 min at 4°C. Cells were washed again with 1% FBS/PBS to remove excess secondary antibody, passed through a cell strainer and run on a FACSAria III (BD Biosciences) to collect the top and bottom 5% of cells expressing surface CD98 into full media.
NCI‐H1299 subcutaneous Xenografts
NCI‐H1299 cells sorted for CD98hi or CD98lo sub‐populations (1 × 104 cells) were injected subcutaneously into the right flank of female Nod‐Scid mice. Groups of 10 mice received subcutaneous inoculation of CD98hi or CD98lo NCI‐H1299 cells. NCI‐H1299 stable cells (1 × 105) expressing stable Lat1, EZH2, or non‐targeting shRNAs were injected subcutaneously into the right flank of female Nod‐Scid mice. Groups of 15 mice received subcutaneous inoculation of pooled Lat1, EZH2, or non‐targeting shRNAs expressing NCI‐H1299 cells. All animals were completely dissected and photographically documented. In vivo tumor growth was measured weekly using standard caliper measurements and statistically analyzed using repeated‐measures ANOVA.
Amino acid HPLC, SAM assay, and metabolomic LC‐MS
Amino acid analysis was completed on a Biochrom 30+ amino acid analyzer as per manufacturer's instructions. Briefly, cells were grown on a 6‐well plate, rinsed with ice‐cold PBS, and scraped into 60 μl of ice‐cold PBS. The scraped cell suspension was disrupted in water bath sonicator, and cellular debris removed by 4°C, 10 min centrifugation at 21,000 g. Fifteen microliters of the resulting cellular lysate was used as injection volume to analyze and compare to amino acid standards.
Cellular SAM was assayed using the Bridge‐It S‐adenosylmethionine cellular assay kit (Mediomics) as per the manufacturer's instructions. Briefly, cells grown in a 6‐well plate were rinsed with ice‐cold PBS and then lysed in 150 μl ice‐cold 0.2% perchloric acid and 0.08% β‐mercaptoethanol. Ten microliters of the resulting lysate was combined with assay reagents, incubated for 60 min at room temperature, transferred to a non‐binding black bottom 96‐well plate, and assayed on a Victor3V plate reader (PerkinElmer) with 495 nm excitation and 665 nm emission settings.
Metabolomic LC‐MS was performed on cells grown on a 6‐well plate 24 h after FACs sorting for CD98hi and CD98lo populations. Cell pellets were rinsed briefly with ice‐cold PBS and extracted with 200 μl cold MeCN/MeOH/H2O 40/40/20 for 10 min, followed by extraction with 300 μl cold water for 5 min. Ten microliters of 2 μM iso‐ATP and 5‐fluoro‐2‐methylpyridine (injection standards) was added to the pooled metabolites, and the resulting solutions were centrifuged at 15,000 × g at 4°C for 15 min. Thirty microliters of the extracts were loaded onto an Scherzo SM‐C18 150 × 2.0 mm column (Imtakt USA) at a flow rate of 0.3 ml/min and column oven temperature of 30°C. The mobile phases used for metabolite separations were water containing 0.1% formic acid (mobile phase A) and acetonitrile containing 0.1% formic acid (mobile phase B). The following gradient was used: 0–6 min: 0% B, 6–16 min 0–50% B, 16–18 min 50–100% B, 18–20 min 100% B, 20–20.5 min 100–0% B, and 20.5–30 min 0% B. The column effluent was delivered directly to an API 4000 mass spectrometer (AB SCIEX) with a heated turbo electrospray ionization source operated in multiple reaction monitoring (MRM) mode.
Gene expression analysis by qRT–PCR
One‐Step quantitative RT–PCR was performed on 100 ng of total RNA from NCI‐H1299 cells using Superscript III Platinum One‐Step qRT–PCR System (Invitrogen) according to the manufacture's protocol. Gene expressions of interest were measured using TaqMan Gene Expression Assays (Applied Biosciences) on the CFX96 Real‐Time System (BioRad). Results were represented as fold change using the ∆∆Ct method, following normalization to RPN1. TaqMan primer/probe sets used in this study are found in Supplementary Table S1.
Chromatin immunoprecipitation‐quantitative PCR (ChIP‐qPCR)
ChIP‐qPCR was carried out following the protocol outlined in De Abrew et al (2010). Briefly, NCI‐H1299 cells were fixed with 1% formaldehyde for 15 min and quenched with 0.125 M glycine. Chromatin was isolated by adding lysis buffer, followed by disruption with a Dounce homogenizer. Lysates were sonicated, and the DNA sheared to an average length of 300–500 bp. Genomic DNA was prepared by treating chromatin with RNase, proteinase K and heated for de‐crosslinking, followed by ethanol precipitation. Pellets were resuspended, and the resulting DNA was quantified on a NanoDrop spectrophotometer. The total chromatin yield was determined by extrapolation to the original chromatin volume.
Purified chromatin (20–30 μg) was precleared with protein G agarose beads (Invitrogen). Genomic DNA regions of interest were immunoprecipitated using the antibodies indicated (Supplementary Table S2). After incubation at 4°C overnight, protein G agarose beads were used to isolate the immune complexes followed by washes, elution from the beads with SDS buffer, and RNase and proteinase K treatment. Crosslinks were reversed by incubation overnight at 65°C, and ChIP DNA was purified by phenol–chloroform extraction and ethanol precipitation.
qPCRs were carried out in triplicate on specific genomic regions using SYBR Green Supermix (Bio‐Rad). The resulting signals were normalized for primer efficiency by carrying out qPCR for each primer pair using Input DNA. The sequences of ChIP‐qPCR primers are listed in Supplementary Table S1.
Stable cell line engineering
All lentiviral‐delivered constitutive shRNAs were purchased from Sigma as high titer virus, and all shRNA sequences are found in Supplementary Table S1. Stable NCI‐H1299 cells were created by exposing cells at 50% confluency in a 6‐well plate to pseudotyped lentiviral particles at an MOI of one for a 24‐h period in the presence of 8 μg/ml polybrene. After 24 h, media was replaced with full media and cells were grown to confluence and then passaged into T‐175 culture flasks with puromycin selection (100 μg/ml) until resistant cell colonies formed and antibiotic induced cell death ceased. Several hundred resistant colonies were pooled and expanded to form the stable cell lines employed herein.
SV40 large‐T antigen was delivered to NHBE as pseudotyped lentiviral particles at an MOI of one (Capital Biosciences). NHBE with large‐T antigen were passaged 20 times beyond the point of telomeric crisis occurring in NHBE without large‐T antigen.
Lentiviral‐delivered inducible MAT2A shRNAs were purchased as DNA constructs from Cellecta, Inc. Lentivirus were packaged as per manufacturer's instructions, and stable cell lines were generated as described above.
Immunoblotting, chromatin immunoprecipitation, and real‐time PCR
Standard Western blotting techniques were used for all immunoblotting experiments described. Cell lysates were generated from 2D and spheroid growth conditions using RIPA buffer containing 40 mM HEPES pH 7.5, 120 mM NaCl, 1 mM EDTA, 5% glycerol, 1% NP40, 1% Na deoxycholate, 0.1% SDS, protease inhibitor cocktail (Sigma), and phosphatase inhibitor cocktails I and III (Sigma). Antibodies used in this study are found in Supplementary Table S2.
Microscopy and immunohistochemistry
Micrographs of live cells grown in 2D and spheroid cell culture were taken using phase contrast settings on a Zeiss Axioscope at the indicated magnifications.
Immunohistochemistry was performed on formalin‐fixed paraffin‐embedded tissue sections using standard protocols and developed via DAB detection and horseradish peroxidase‐conjugated secondary antibody. Tissue microarrays with both normal and tumor tissue were procured from US Biomax. Slides were counterstained using standard hematoxylin staining protocol.
SGD completed and designed experiments and/or data collection included in all Figures. MR completed the majority of experiments presented in Fig 3. AMB completed the experiment presented in Fig 3C. JG developed and applied immunohistochemistry assays presented in Fig 4. CS completed live cell sorting experiments presented in Fig 1. MM completed the matrigel growth assay presented in Fig 5H. CH, LL, JL and SJG completed the in vivo assays presented in Fig 1D, 5I, and Supplementary Fig S4C. MK completed the statistical analysis from immunohistochemical staining of lung cancer tissue microarrays. FW and JSM completed the metabolomic profiling and bioinformatic analysis presented in Fig 1E. KGG completed the experiment presented in Fig 2C. MTF contributed the GeneLogic mRNA expression profiling data presented in Fig 4A and Supplementary Fig S3. AK, RAR and VRF contributed experimental design and data interpretation essential in completing the manuscript.
Conflict of interest
All listed authors are or were full‐time employees of Pfizer, Inc. with no relationships that they believe could be construed as resulting in an actual, potential, or perceived conflict of interest with regard to the manuscript submitted for review.
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Figure Legends and Supplementary Tables
- © 2015 The Authors