Valproic acid targets IDH1 mutants through alteration of lipid metabolism
IDH1 MT glioma are sensitive to VPA in vitro and in vivo
We tested three different HDACis, VPA, panobinostat, and belinostat in varying concentrations in IDH1 WT and IDH1 MT primary glioma cell lines. IDH1 MT glioma cell lines had a slightly lower IC50 for VPA compared to IDH1 WT cell lines (WTIC50 = 1.23; MTIC50 = 0.96), and the hill slope between the two dose–response curves were statistically significant (WTHill Slope = −0.80; MTHill Slope = −1.68; 0.0032) (Fig. 1A). Panobinostat (Fig. 1B) inhibited the growth of both IDH1 WT and IDH1 MT cell lines but there was no significant difference in the IC50 or hill slope between the two dose–response curves. The IDH1 WT cell lines tested were more sensitive to belinostat than IDH1 MT glioma cell lines (WTIC50 = 0.84; MTIC50 = 1.97) (Fig. 1C). As expected, the heterozygous IDH1 MT cell lines showed significantly higher 2HG levels compared to an IDH1 WT cell line (Fig. 1D). The 2HG content in hemizygous IDH1 MT cell line BT142 was approximately double that of the IDH1 WT cell line but the difference did not reach statistical significance. The heterozygous IDH1 MT cell lines HK252 and HK211 do not form xenograft in vivo, even in NSG mice; hence we tested the efficacy of VPA in vivo in the hemizygous BT142 model. Although BT142 has limited 2HG production (Fig. 1D), we found that the transcriptome of BT142 is most similar to our heterozygous IDH1 MT cell line HK252 and still distinct from IDH1 WT cell lines (Supplementary Fig. 1A), indicating that even with the loss of 2HG production, the line possesses many of the essential features of IDH1 mutant gliomas. We found that in vivo VPA treatment significantly improved the survival of mice in the BT142 xenograft model by 17 days (Fig. 1E).
Next, we tested the efficacy of HDACi’s in a syngeneic murine IDH1 (mIDH1) glioma model13. In vitro, NPAIC1(NRAS/IDH11R132H/shP53/shATRX) compared to mIDH1 WT cell line NPAC54B (NRAS/shP53/shATRX) had a slightly lower IC50 for both VPA (NPAC54BIC50 = 1.65; NPAIC1IC50 = 1.37) (Fig. 2A) and belinostat (NPAC54BIC50 = 0.57; NPAIC1IC50 = 0.28) (Fig. 2B). The mIDH1 MT cell line NPAIC1 also had significantly higher 2HG than mIDH1 WT cell line NPAC54B (Fig. 2C).
In vivo, VPA treatment did not improve the survival of mice bearing mIDH1 WT tumor xenografts (Fig. 2D) but significantly improved the survival of mice bearing mIDH1 MT tumor xenografts by 10 days (Fig. 2E). Thus, in the tumor models that we examined, we established that VPA diminished tumor growth in vivo and in vitro, but the murine model suggests a greater efficacy in IDH1 mutant tumors.
VPA has opposing effects on overall gene expression and chromatin accessibility
VPA is a multifaceted drug that has several direct14 and indirect targets15. Given the pleiotropic effects of VPA, we wanted to understand why IDH1 MT glioma cell lines showed some selectivity to VPA treatment. Given that HDACs are a known target of VPA, we treated an IDH1 MT cell line HK252 and an IDH1 WT cell line HK157 with 1 mM VPA and examined histone 3 lysine 27 acetylation (H3K27ac) over time. We found that after 96 h of treatment, VPA induced a strong increase in H3K27ac in both IDH1 WT (Fig. 3A) and IDH1 MT (Fig. 3B) cell lines, as would be expected with HDAC inhibition. We next treated multiple IDH1 MT glioma cell lines, HK252, HK213, BT142, & MGG119 (4) and IDH1 WT glioma cell lines, HK157, HK408, HK206, HK350, & HK393 (5) with 1 mM VPA for 48 h and conducted bulk RNA sequencing to get a better understanding of the transcriptomic effect of VPA. A greater number of genes were upregulated than downregulated with VPA treatment in both IDH1 WT and IDH1 MT glioma cell lines (Fig. 3C). The directionality of individual gene expression changes was similar for most WT and MT cell lines tested. However, we did notice some variability in the magnitude of individual gene expression changes among individual MT and WT cell lines tested. To better understand what biological processes are altered by VPA, we performed gene set enrichment analysis (GSEA) and compared the gene ontology (GO) terms that were up and downregulated with VPA treatment in IDH1 WT and IDH1 MT cell lines. We found in both IDH1 WT and IDH1 MT primary glioma cell lines, GO terms related to cell cycle and DNA repair were downregulated, whereas GO terms related to neurogenesis were upregulated after VPA treatment (Fig. 3D). We next asked what biological processes were only altered in IDH1 MT and not in IDH1 WT after VPA treatment. We found that VPA upregulated biological processes related to differentiation and downregulated lipid and sterol biosynthetic processes in the IDH1 MT but not IDH1 WT cell lines (Fig. 3E). We examined a panel of lipogenic genes and found that SREBF1, ACLY, FASN, SCD, and HMGCR which are all part of the de novo lipogenesis pathway were significantly downregulated in primary IDH1 MT cell lines after VPA treatment (Fig. 3F). In the IDH1 MT cell lines VPA inhibited expression of FASN and HMGCR more significantly than in the IDH1 WT cell lines. We also treated mIDH1WT and mIDH1MT with 1 mM VPA for 48 h and conducted RNA sequencing. We found that VPA downregulated FASN in the mIDH1MT but not in the mIDH1WT cell line (Supplementary Fig. 2A).
We next re-analyzed our previously published RNA-seq and ATAC-seq dataset16 of IDH1 MT and IDH1 WT cell lines treated with 1 mM VPA for 5 days and found that similar to the current findings, VPA treatment resulted in an increase in the expression of a greater number of genes and downregulated a smaller number of genes in both IDH1 WT and IDH1 MT cell lines (Fig. 3G). Interestingly, ATAC seq analysis showed that despite the relatively increased transcription, a greater number of peaks were lost, and only a few peaks were gained after VPA treatment (Fig. 3H), suggesting that VPA treatment activates transcription but paradoxically results in chromatin condensation. We next integrated ATAC and RNA seq data and when we examined the promoter region of VPA-upregulated genes in both IDH1 WT and IDH1 MT cell lines, we found that most genes did not show a change in promoter accessibility (Fig. 3I). There was a stronger correlation between promoter chromatin accessibility and VPA downregulated genes in IDH1 MT compared to the IDH1 WT cell lines (Fig. 3J). Interestingly, promoter accessibility of lipogenic genes was significantly altered in the IDH1 MT but not IDH1 WT cell lines (Fig. 3K). Taken together, our findings thus far indicate that VPA, while inhibiting the growth of both IDH1 mutant and wildtype cells, has different molecular effects on the IDH1 mutant and wildtype cells that we examined.
The mTOR pathway is involved in the downregulation of FASN by VPA
The mechanistic target of rapamycin (mTOR) signaling pathway plays an important role in regulating lipogenesis17 and VPA has been found to inhibit the mTOR pathway in some cancer cell lines18,19, so we wondered whether VPA might be inhibiting lipogenic genes via the mTOR pathway. FASN is a key enzyme in the de novo lipogenesis pathway and although VPA targeted several enzymes involved in lipid metabolism, VPA inhibited transcription of FASN significantly more in the IDH1 MT compared to IDH1 WT. In addition, both in our IDH1 MT primary tumor cell lines and murine model, VPA inhibited transcription of FASN; hence we focused our attention on FASN.
We treated IDH1 WT cell line HK157 and 2 IDH1 MT glioma cell lines, HK252 and BT142, with VPA for four days and examined phosphorylation of ribosomal protein S6 (PS6) and FASN protein expression. While VPA treatment had a minimal effect on PS6 and FASN protein expression in the IDH1 WT cell line HK157 (Fig. 4A) it resulted in diminished PS6 and FASN protein expression in IDH1 MT cell lines HK252 (Fig. 4B) and BT142 (Fig. 4C). Treatment of HK252 with VPA resulted in the reduction of pS6 within 24 h, possibly as early as 4 h (Fig. 4D). However, we note a reduction in FASN within 2 h. This suggests that the initial reduction of FASN expression is not directly downstream of mTOR inhibition. Treatment of the IDH1 MT cell line HK252 with rapamycin (RAPA) also diminished PS6 and FASN protein expression (Fig. 4E). This is consistent with previous literature that found that rapamycin treatment can contribute to reduced FASN expression20. In HK252, the combination of VPA with RAPA was not better than RAPA alone in decreasing FASN protein expression (Fig. 4F). RAPA treatment also inhibited FASN mRNA expression in HK252 while the addition of VPA to RAPA did not further diminish mRNA expression of FASN in HK252 (Fig. 4G). In BT142 VPA concentration greater than 1 mM was better at suppressing PS6 and FASN (Fig. 4C), and the combination of VPA1mM with RAPA100nM was better at suppressing FASN protein expression (Fig. 4H). Interestingly, the combination of VPA with RAPA was not better than VPA alone in inhibiting growth in HK252, but in HK157 and BT142, the combination of VPA and RAPA was better than VPA alone in inhibiting growth (Fig. 4I). This suggests that the combination of VPA with RAPA may be better in growth inhibition for some IDH1 MT cell lines due to better FASN suppression.
We also tested the effect of VPA and RAPA on mIDH1 WT and mIDH1MT cell lines. Surprisingly, VPA treatment resulted in increased PS6 in the mIDH1WT cell line but treatment diminished PS6 in the mIDH1 MT cell line in a dose-dependent manner (Supplementary Fig. 3A). Treatment with RAPA inhibited the growth of only mIDH1 MT but not mIDH1 WT. The combination of VPA and RAPA was better at inhibiting the growth of mIDH1 WT but not mIDH1 MT cell line (Supplementary Fig. 3B). In summary, we conclude that in both the IDH1 MT human primary tumor cell lines and in the murine model VPA inhibits growth at least partially via inhibition of the mTOR pathway and that the mTOR pathway is involved in downregulation of FASN. However, there is line-to-line heterogeneity in the relative extent to which FASN and mTOR signaling are connected.
Inhibition of FASN inhibits the growth of IDH1 MT glioma cell lines
Next, to test whether direct inhibition of FASN has any effect on the growth of IDH1 MT cell lines, we treated IDH1 WT and MT human gliomashpere cell lines with a selective FASN inhibitor (FASNi) TVB-2640 (Fig. 5A). We found that TVB-2640 inhibited the growth of WT cell line HK157 (Fig. 5B) but not HK217 (Fig. 5C) and inhibited growth of both MT cell lines HK252 (Fig. 5D) and BT142 (Fig. 5E). The MT cell lines tested on average were more sensitive to TVB-2640 than the WT cell lines tested (Fig. 5F). Supplementation of palmitate in the media rescued the growth inhibitory effect of TVB-2640 in both WT and MT cell lines, indicating that the negative effects on growth were likely mediated by FASN inhibition. In the murine model, TVB-2640 inhibited the growth of both mIDH1 WT and mIDH1 MT(Supplementary Fig. 4A). Thus, while FASN expression is more sensitive to VPA treatment in IDH1 mutant cells, both IDH1 MT and IDH1 WT cells are sensitive to direct FASN inhibition.
To better understand the effects of FASN inhibition, we treated IDH1 MT gliomasphere cultures with TVB-2640, palmitate, and the combination for 4 days and performed RNA sequencing. We found that TVB-2640 downregulated biological processes related to DNA repair and cell cycle in both HK252 (Fig. 5G) and BT142 (Fig. 5H). Genes related to cell cycle, such as (E2F1, E2F2, RRM2) and anti-apoptosis (DDIAS, MCM10), were downregulated by TVB-2640, and the expression of these genes could be recovered by palmitate supplementation (Fig. 5I). However, we noticed that TVB-2640 upregulated different biological processes in HK252 and BT142. In HK252, biological processes such as TGFβ production, stress fiber assembly and unfolded protein response were upregulated, whereas in BT142, sphingolipid and cholesterol biosynthetic processes were upregulated after FASN inhibition. When we examined individual genes, we found that 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR) was upregulated in both HK252 and BT142. Treatment with lovastatin, an HMGCR inhibitor, inhibited the growth of both MT cell lines HK252 and BT142 (Supplementary Fig. 5A). The combination of lovastatin with TVB-2640 was better at inhibiting growth of HK252 (Supplementary Fig. 5B). This suggests that IDH1 MT glioma cell lines perhaps increase cholesterol biosynthesis as a compensation for FASN inhibition21. When we compared genes that were targeted by both TVB-2640 and VPA we found a suite of genes such as E2F1, E2F2, RRM2, DDIAS, and MCM10 were downregulated by both drugs. In BT142, these genes decreased after 5 days of treatment. However, VPA inhibited the expression of HMGCR, while TVB-2640 upregulated HMGCR in IDH1 MT cell lines. Together, we conclude that in IDH1 MT glioma cell lines, VPA may work through FASN to inhibit cell cycle and anti-apoptotic genes. However, while VPA treatment inhibited transcription of several lipogenic genes, TVB-2640 treatment did not inhibit all lipogenic genes but upregulated some lipogenic genes such as HMGCR and ACACA, suggesting the net metabolic changes induced by VPA and TVB-2640 may be different.
We next assessed the combinatorial effect of TVB-2640 and VPA. We treated HK252 (Fig. 5J) and BT142 (Fig. 5K) with VPA and TVB-2640 and the combination of both drugs for 1 week. In HK252, the combination of VPA with TVB-2640 was not significantly better than TVB-2640 (1 µM) alone in inhibiting growth. In BT142, the combination of VPA2mM with TVB-2640 (1 µM) but not the combination of VPA (1 mM) with TVB-2640 (1 µM) was significantly better than TVB-2640 (1 µM) alone in inhibiting growth. The lack of additivity of the two drugs suggests that at least part of the effect of VPA is mediated via FASN.
VPA and TVB-2640 have distinct effects on free fatty acids in an IDH1 MT glioma cell line
The primary product of FASN is palmitate. Therefore, we next conducted lipidomics to measure palmitate and other free fatty acids after VPA orTVB-2640 treatment. In the IDH1 WT gliomasphere cell line, HK157, treatment with both VPA (Fig. 6A) and TVB-2640 (Fig. 6B) for 4 days significantly decreased free palmitic and stearic acid. Surprisingly, treatment with TVB-2640 significantly increased free palmitic and stearic acid in HK252 (Fig. 6C), suggesting the involvement of a compensatory mechanism or mechanisms. VPA treatment, on the other hand, significantly increased oleic acid but did not alter palmitic or stearic acid in the IDH1 MT cell line HK252 (Fig. 6D). This further suggests that while VPA targets FASN, the net metabolic effect of VPA and TVB-2640 on the lipidome is distinct in the IDH1 MT glioma cell line and it is different when compared to the IDH1 WT cell line HK157. In the IDH1 WT cell line HK157, both VPA and TVB-2640 decreased free saturated fatty acids, while in HK252 TVB-2640 significantly increased free saturated fatty acids. However, in HK252 VPA shifted the ratio of free fatty acids towards monounsaturated fatty acids, suggesting that VPA may not be solely working via inhibition of FASN.
VPA and TVB-2640 alter lipid droplets and induce apoptosis
To prevent lipotoxicity, excess free fatty acids are often converted into neutral lipids and stored in organelles called lipid droplets. Both oleic acid22 and VPA23 have been previously found to induce lipid droplets. Our lipidomic data suggests that both VPA and TVB-2640 can increase free fatty acids in IDH1 MT, so we wondered if this might also have an effect on lipid droplet formation. In the IDH1 MT cell line HK252, treatment with VPA for 4 days increased the number of lipid droplets in lipid droplet-positive cells, as shown in Fig. 7B and quantified in Fig. 7I when compared to control (Fig. 7A, I). Treatment with TVB-2640 did not increase lipid droplet formation (Fig. 7C, I). This is probably because treatment with TVB-2640 alone does not significantly increase oleic acid, which is a lipid droplet inducer22. The combination of VPA with TVB-2640 inhibited the accumulation of lipid droplets (Fig. 7D, I), perhaps by preventing a significant increase in free oleic acid. Treatment of the IDH1 MT cell line HK252 with 100 µM oleic acid also increased lipid droplets (Fig. 7E, K), and TVB-2640 in combination with oleic acid (Fig. 7F, K) inhibited lipid droplet formation. The combination of VPA and oleic acid also increased lipid droplets compared to control (Fig. 7K). Lipid droplets in the VPA with oleic acid condition appeared much larger and appeared to fuse with each other (Fig. 7G). Interestingly, unlike HK252, where we found a significant increase in lipid droplets with both VPA (Fig. 7I) and oleic acid (Fig. 7K), in HK157, oleic acid significantly increased lipid droplets (Fig. 7J), but VPA did not (Fig. 7L). These findings indicate that the increase in lipid droplets in HK252 may be connected to the increase in free oleic acid induced by VPA. However, in HK157, we saw a decrease in free fatty acids, and hence we saw no change in lipid droplets.
We further quantified the amount of annexin V positive cells after VPA, TVB-2640, and oleic acid treatment and found that, compared to control, all three treatments did not increase the proportion of increased the proportion of annexin V positive cells in the IDH1 WT cell line, HK157 (Fig. 7L) but did in the only in the IDH1 MT cell line HK252 (Fig. 7M). Oleic acid treatment also inhibited the growth of IDH1 MT cell line HK252 but not WT cell line HK157 in a dose-dependent manner (Fig. 7N), indicating that one mechanism of cell death in VPA treatment could be lipotoxicity due to oleic acid. We also examined what fraction of cells that were positive for lipid droplets were also positive for annexin V by flow cytometry (Supplementary Fig. 6A). In MT cell line HK252, we found that after VPA treatment about 30% of the annexin V positive cells were also positive for lipid droplets whereas 70% of annexin V positive cells were not positive for lipid droplets (Supplementary Fig. 6B). This suggests that not all cells that are treated with VPA undergo lipotoxicity-induced apoptotic cell death. The formation of lipid droplets in a fraction of apoptotic cells (30%) may be a compensatory mechanism to protect cells from lipotoxicity, although one that is incomplete in that the cells still undergo apoptosis.
Interestingly, TVB-2640, in combination with VPA, prevented lipid droplet formation, suggesting there might be some therapeutic benefit of combining VPA with TVB-2640. Lastly, treatment with oleic acid inhibited PS6 (Fig. 7O) in IDH1 MT cell lines suggesting oleic acid may indirectly contribute to mTOR inhibition by VPA.
HDACs are involved in the regulation of lipogenic gene expression
Next, given the pleiotropic effects of VPA and the lack of selectivity of other broad-spectrum HDACi for IDH1 mutant tumors, we wondered whether the effect of VPA on lipogenic genes is dependent or independent of HDAC inhibition. We re-analyzed RNA seq data from our recently published work16 and assessed the effect of HDAC knockdown on fatty acid metabolism genes. We found that knockdown of HDAC 2, 3, 4, and 9 increased the expression of lipogenic genes such as FASN, stearyl coA desaturase (SCD) and HMGCR. HDAC1 knockdown slightly decreased the expression of a few lipogenic enzymes but interestingly, only HDAC6 knockdown inhibited transcription of FASN. HDAC6 knockdown also inhibited transcription of other lipogenic enzymes such as ACACA, SCD and HMGCR but seemed to have a greater inhibitory effect on FASN (Supplementary Fig. 7A). Using 2 different CRISPR sgRNAs, we prospectively knocked down HDAC6 in the HK252 cell line. HDAC6 is a microtubule deacetylase, and as expected, HDAC6 knockdown increased tubulin acetylation (Supplementary Fig. 7B). Similarly, treatment of HK252 with VPA increased tubulin acetylation (Supplementary Fig. 7C) in a dose-dependent manner suggesting that VPA may mediate some of its effect on lipogenic genes through HDAC6. HDAC6 knockdown also inhibited the transcription of FASN (Supplementary Fig. 7D) and inhibited the growth of MT cell line HK252 but not the WT cell line HK157 (Supplementary Fig. 7E). Taken together, the data indicate that the effect on lipogenic genes by VPA may be mediated through HDAC6. However, the upregulation of some of the lipogenic enzymes after HDAC knockdown potentially suggests a mechanism by which tumor cells might compensate and why some HDAC knockdowns inhibit the growth of IDH1 MT glioma cells and others do not.
FASN knockdown inhibits IDH1 mutant tumor growth in vivo and may enhance the effects of VPA treatment
Because tumor cells may be able to compensate for the loss of FASN function in vivo through the utilization of exogenous fatty acids, it was critical for us to determine the effect of diminished FASN function in vivo. It is currently unclear whether TVB-2640 can cross the blood–brain barrier and the extent of off or on-target toxicities of the drug in mice. Hence, to test the effect of FASN inhibition in vivo we knocked down FASN in IDH1 MT lines BT142 and HK252 (Fig. 8A) using shRNAs. In vitro, FASN knockdown inhibited the growth of both HK252 (Fig. 8B) and BT142 (Fig. 8C). In BT142, palmitate rescued some of the growth inhibitory effect of FASN knockdown with FASN shrna1 (Fig. 8D). We further found that FASN knockdown in combination with VPA was better at inhibiting growth of BT142 when compared to VPA and FASN knockdown alone (Fig. 8D) in vitro. Interestingly, palmitate did not rescue the growth inhibitory of FASN knockdown in combination with VPA. Therefore, we further hypothesized that an in vivo combination of VPA with FASN knockdown may have a better growth inhibitory effect than either alone.
We found in vivo, compared to control, both VPA and FASN knockdown alone improved the survival of mice (Fig. 8E) in the BT142 model. Further, the combination of VPA with FASN knockdown slightly but significantly improved the survival of mice when compared to VPA or FASN knockdown alone (Fig. 8E). These findings indicate that in vivo, even in the presumed presence of exogenous palmitate, FASN knockdown alone can improve survival of mice. However, cancer cells can often use multiple ways to metabolically compensate or resist treatment; hence knockdown of FASN may improve the response to VPA in vivo.
link