Fluoxetine in H460 Lung Cancer

Central Finding

SREBF1/2-driven lipid biosynthesis + MYC post-translational suppression. SREBF1 is the #1 activated transcription factor among 688 tested, and 18 of the top 50 DEGs are lipid biosynthesis genes.

Original Paper's Model

The ATF4-AKT-mTOR axis is partially supported. PERK-ATF4-CHOP is selectively activated, but ATF4 mRNA is unchanged (padj=0.665) — activation is purely post-translational. ATF4 is not the master transcriptional regulator.

Key Methodological Insight

ER stress GO terms are not enriched among upregulated DEGs: "response to ER stress" (padj=0.56) and "unfolded protein response" (padj=0.35). This contradicts the original paper's ORA-based finding and demonstrates how enrichment method choice impacts conclusions.

Data Quality

All 6 samples pass QC with no outliers. 23,853 genes detected after filtering. Within-group Pearson correlations exceed 0.994, confirming excellent replicate reproducibility.

Treatment Effect

The dominant source of transcriptomic variation is fluoxetine treatment (47.4% of total variance on PC1), indicating a strong and consistent drug effect. The systematic lower library sizes in treated samples (mean 16.7M vs 23.2M) may reflect reduced transcriptional output consistent with growth inhibition.

Lipid Gene Dominance

18 of the top 50 DEGs are cholesterol/fatty acid biosynthesis genes (INSIG1, HMGCS1, FADS2, MSMO1, SCD, SQLE, HMGCR) — all controlled by SREBP transcription factors. This is the dominant transcriptomic signal, not ER stress.

ATF4 Not Transcriptionally Changed

ATF4 mRNA: log2FC = -0.108, padj = 0.665 (not significant). The original paper’s central hub is activated purely post-translationally via eIF2α phosphorylation, invisible in transcriptomic data.

Exact Match with Original Paper

Our analysis identified exactly 166 DEGs, matching the original paper’s report of 166 DEGs — strong cross-validation of the computational pipeline.

MYC: The Strongest Signal

MYC_TARGETS_V1 (NES=-3.615) is the single strongest enrichment signal, indicating massive shutdown of MYC-driven transcription affecting ribosome biogenesis, translation, and RNA processing. The original paper never identified MYC suppression.

KEGG ER Stress Failure

KEGG “Protein processing in ER” is NOT significant (NES=0.963, FDR=0.574), contradicting the original paper’s ORA-based claim. The discordance arises because GSEA ranks all genes while ORA only tests the 166 DEGs.

Lysosome & FIASMA

Lysosome (NES=2.869), Sphingolipid metabolism (NES=2.026), and Ferroptosis (NES=2.001) reflect fluoxetine’s cationic amphiphilic drug (CAD) mechanism: lysosomal accumulation, ASM inhibition, and membrane disruption.

SREBF1/2 Confirmed

SREBF1 (diff=1.723, padj=0.024) and SREBF2 (diff=1.563, padj=0.024) are the #1 and #2 most activated TFs among 688 tested. ATF4 (diff=0.123, padj=0.456) is not significant — confirming SREBP, not ATF4, as the master regulator.

Lipid-Sensing TF Module

SREBF1/2 + PPARα (diff=0.639) + LXRα/NR1H3 (diff=0.634) + ChREBP (diff=0.461) form a coherent lipid-sensing regulatory network — both synthesis and catabolism TFs activated simultaneously.

MAPK → MYC Cascade

MAPK pathway repressed (diff=-0.881, padj=0.018) connects to MYC target suppression (GSEA NES=-3.615). MAPK activates MYC via Ser62 phosphorylation — its repression leads to MYC protein degradation and massive proliferative gene shutdown.

TNFα Discordance Resolved

PROGENy shows TNFα REPRESSED (diff=-0.265) despite GSEA showing positive enrichment (NES=1.753). NF-κB TF activity is zero (NFKB1 diff=0.001). The hallmark gene set is activated by stress-responsive TFs, not TNFα signaling.

SERT Absent

SLC6A4/SERT is expressed at only 0.13 CPM — far below the ~1 CPM functional threshold for protein expression. Serotonin synthesis enzymes (TPH1 0.12 CPM, TPH2 0.04 CPM, DDC 0.10 CPM) are absent. H460 cells lack the capacity for serotonin-dependent signaling. Fluoxetine's anti-cancer effects are entirely off-target, mediated by CAD/FIASMA properties (lysosomal accumulation, ASM inhibition, membrane lipid disruption).

Kynurenine Pathway

KYNU (kynureninase) is highly expressed (428.94 CPM) and significantly upregulated (log2FC=0.300, padj=0.010), though below the |log2FC|>1 DEG threshold. Since IDO1/IDO2 are absent, the kynurenine pathway operates via TDO2 (14.07 CPM). KYNU upregulation may affect NAD+ biosynthesis and production of immunomodulatory metabolites (kynurenic acid, quinolinic acid).

PERK-Dominant Pattern

The PERK→eIF2α→ATF4→CHOP axis is the only UPR branch with clear transcriptomic activation. PERK branch has 4/7 genes nominally significant vs 0/8 for ATF6. This selective PERK activation is characteristic of metabolic/lipid ER stress, not the protein misfolding that would engage ATF6 chaperone expansion. The functional output is visible through ATF4's transcriptional targets: DDIT3/CHOP and TRIB3 (both padj<0.05), with CHAC1 trending upward (padj not computable due to low counts).

ATF4 Post-Translational

ATF4 mRNA is unchanged (log2FC=-0.11, padj=0.665), yet its transcriptional targets (DDIT3, TRIB3, ATF3, WARS1) are all upregulated. ATF4 is activated via eIF2α phosphorylation which derepresses ATF4 translation from upstream ORFs — a purely post-translational mechanism. This explains why the original paper's Western blot detected ATF4 protein upregulation while our RNA-seq shows no mRNA change.

PERK mRNA Feedback

EIF2AK3/PERK kinase is transcriptionally upregulated (log2FC=0.88, padj=8.7e-8) — the most statistically significant change in the entire UPR panel, though below DEG fold-change threshold. This novel finding suggests a positive transcriptional feedback loop amplifying PERK signaling capacity. Paradoxically, its substrate EIF2S1/eIF2α is downregulated (log2FC=-0.33, padj=0.008), possibly reflecting a compensatory mechanism.

Autophagy Dominant

SQSTM1/p62 (padj=1.73e-21), OPTN (padj=1.83e-13), and BNIP3L (padj=1.95e-12) are among the most significantly changed genes in the entire transcriptome. OPTN and BNIP3L are selective autophagy/mitophagy receptors, implicating targeted removal of damaged mitochondria. ULK1 (initiation), MAP1LC3B (autophagosome formation), and LAMP1/2 (lysosomal fusion) are all upregulated, showing activation across the full autophagy pathway.

Apoptotic Priming

BCL2L1/BCL-XL is downregulated (log2FC=-0.69, padj=1.28e-5), reducing anti-apoptotic defense. Yet BAX, BAK1, CASP3, and CASP9 mRNAs are unchanged — confirming the original paper's caspase-3 cleavage is entirely post-translational. The transcriptome shows "apoptotic priming": cells lower their defenses (BCL-XL down, BID down) while actual execution occurs at the protein level via proteolytic caspase cascades.

Ferroptosis: Armed but Defended

ACSL4 (padj=1.55e-6) incorporates PUFAs into membranes, sensitizing cells to lipid peroxidation. HMOX1 (padj=7.66e-9) releases free iron from heme. Both are pro-ferroptotic. However, SLC7A11/xCT (padj=0.040) imports cystine for glutathione synthesis, and FTL (padj=0.016) sequesters iron — both protective. GPX4 (the key ferroptosis suppressor) is unchanged. Cells mount an active transcriptional defense against ferroptotic stress from fluoxetine's lipid disruption.

Ceramide Clearance Triad

Three distinct ceramide clearance pathways are simultaneously activated: ASAH1 (hydrolysis, padj=4.1e-27), UGCG (glycosylation, padj=1.4e-15), and CERT1 (ER→Golgi transport, padj=2.9e-10). This coordinated multi-pathway response suggests fluoxetine creates sufficient ceramide stress to engage every available cellular mechanism for ceramide disposal. UGCG upregulation is particularly notable as it is associated with multidrug resistance in cancer.

FIASMA Compensatory Response

SMPD1/ASM is nominally upregulated (log2FC=0.35, padj=0.037). Fluoxetine inhibits ASM post-translationally by displacing the enzyme from the inner lysosomal membrane. The transcriptional upregulation represents a compensatory response — cells increase SMPD1 mRNA to offset functional inhibition. This is the first transcriptomic evidence of ASM compensatory transcription in response to a FIASMA in cancer cells.

DHCer → Autophagy Link

DEGS1 (dihydroceramide desaturase) is downregulated (padj=0.025) while de novo synthesis is enhanced (SPTLC2, KDSR up). This creates a metabolic funnel: more sphingoid base synthesis flowing into dihydroceramide, which cannot be efficiently converted to ceramide. Dihydroceramide accumulation has been independently linked to autophagy induction, connecting this pathway directly to the dominant autophagy signal in cell death programs.

Non-Immunogenic Death

HMGB1 is significantly downregulated (padj=1.6e-4) — this DAMP alarmin is normally released during immunogenic cell death to activate dendritic cells. Its suppression, combined with CCL2 downregulation (reduced monocyte recruitment) and complete NF-kB inactivity (NFKB1/RELA unchanged, A20 significantly up), indicates fluoxetine-induced death is immunosuppressive. This has direct implications for combination therapy: immune checkpoint inhibitors may not synergize well if the death mode is non-immunogenic.

Non-Genotoxic G1 Arrest

All DDR sensors and checkpoint kinases (ATM, ATR, CHEK1, CHEK2, TP53) are transcriptionally unchanged — fluoxetine does not cause DNA damage. Cell cycle arrest is achieved through post-translational stabilization of p21/p27 (mRNA unchanged, padj>0.35) and passive suppression of S/G2/M progression genes. The most significant changes are in G2/M regulators (7/8 nominally significant), with CDC20 (padj=2.95e-11) near the DEG threshold (log2FC=-0.907). E2F1 downregulation (padj=1.13e-4) reflects autoregulatory feedback from reduced proliferation.

Post-Translational MYC

MYC mRNA is completely unchanged (log2FC=0.051, padj=0.852), yet MYC_TARGETS_V1 is the strongest hallmark signal (NES=-3.615). This definitively proves MYC suppression occurs at the protein level, consistent with MAPK-mediated MYC phosphorylation and proteasomal degradation (MAPK pathway repressed, PROGENy padj=0.018). The MAPK→MYC post-translational cascade explains the massive S/G2/M gene suppression without changes to upstream transcription factors or CDK inhibitors.

Unified Model

Fluoxetine acts as a CAD/FIASMA agent, disrupting lysosomal membranes and triggering a coordinated program: SREBF1/2-driven lipid biosynthesis + MYC post-translational suppression. MAPK repression (padj=0.018) drives post-translational MYC degradation, while three independent ceramide clearance pathways (ASAH1, UGCG, CERT1) manage the sphingolipid disruption.

Original Paper Reassessed

The ATF4-AKT-mTOR axis model is partially supported but substantially incomplete. ATF4 mRNA is unchanged (padj=0.665) — activation is purely post-translational. SREBF1/2 are the true master transcriptional regulators (SREBF1 #1 of 688 TFs, diff=1.723).

Therapeutic Implication

ASAH1 inhibition (e.g., carmofur) could synergize with fluoxetine by blocking the ceramide clearance escape route. Cell death is likely non-immunogenic (HMGB1↓, CCL2↓, NF-kB inactive), suggesting that immune checkpoint inhibitors may have limited efficacy alongside fluoxetine in this context.