Proteomic profiling reveals ACSS2 facilitating metabolic support in acute myeloid leukemia

Acute myeloid leukemia (AML) is a heterogeneous disease characterized by genomic aberrations in oncogenes, cytogenetic abnormalities, and an aberrant epigenetic landscape. Nearly 50% of AML cases will relapse with current treatment. A major source of therapy resistance is the interaction of mesenchymal stroma with leukemic cells resulting in therapeutic protection. We aimed to determine pro-survival/anti-apoptotic protein networks involved in the stroma protection of leukemic cells. Proteomic profiling of cultured primary AML (n = 14) with Hs5 stroma cell line uncovered an up-regulation of energy-favorable metabolic proteins. Next, we modulated stroma-induced drug resistance with an epigenetic drug library, resulting in reduced apoptosis with histone deacetylase inhibitor (HDACi) treatment versus other epigenetic modifying compounds. Quantitative phosphoproteomic probing of this effect further revealed a metabolic-enriched phosphoproteome including significant up-regulation of acetyl-coenzyme A synthetase (ACSS2, S30) in leukemia-stroma HDACi treated cocultures compared with untreated monocultures. Validating these findings, we show ACSS2 substrate, acetate, promotes leukemic proliferation, ACSS2 knockout in leukemia cells inhibits leukemic proliferation and ACSS2 knockout in the stroma impairs leukemic metabolic fitness. Finally, we identify ACSS1/ACSS2-high expression AML subtype correlating with poor overall survival. Collectively, this study uncovers the leukemia-stroma phosphoproteome emphasizing a role for ACSS2 in mediating AML growth and drug resistance.

parameters were left as default.Data quality was inspected using the in-house developed tool PTXQC (25).A correlation plot indicated a low variation of protein expression across all quality control samples, thus no normalization was required before data analysis.A total of 6754 proteins were quantitated.A filter was applied to retain proteins having LFQ intensity values in at least 14 of the 28 samples, yielding a total of 2231 unique proteins.Differential protein expression between AML Mono and AML Cocu was determined using the limma package in the statistical language R, requiring a fold change ≥1.5 and <0.5, q-value ≤0.05.

KG1a, Hs5, and KG1a-Hs5
Samples were analyzed using a QExactive mass spectrometer (Thermo Scientific) and a U3000 RSLC nano HPLC system (Dionex) coupled to a 50 cm-long and 75micron internal diameter EASYspray column.Samples were resolved using a 4 h gradient with the mass spectrometer run in data-dependent analysis mode, in which the 20 most intense multiply charged precursors were selected for fragmentation.The mass spectrometer settings were as follows: Precursor resolution; 70000, AGC target; 3000000, maximum fill time; 250 ms, MSMS resolution; 17500, AGC target 100000' maximum fill time was 120 ms, isolation window; 3 Da, normalized collision energy (NCE); 30, underfill ratio 10%.Data were processed and quantified using Proteome Discoverer 1.4 (Thermo Scientific) and Mascot (MatrixScience).Search parameters in Mascot (via PD) were as follows -Enzyme (trypsin), Dynamic modifications: Oxidation (M), Phosphorylation  K=2 and 30k iterations).Correlations between continuous factors of specified groups of patients were calculated with Kruskal-Wallis rank sum test, while dichotomous factors were tested with Pearson's chi-squared, both using the 'tableBy' function in the 'arsenal' package in R.
To study the impact of the metabolic gene expression on patient survival, the overall survival (in months, mo) was performed using the R package 'survival', with times censored at the last follow-up.Differences between groups were tested with unadjusted Kaplan-Meier curves using log-rank tests.Hazard ratios were estimated from Cox proportional hazard regression and used to evaluate the independent effects of covariates.Univariate and multivariate analyses were evaluated for Cox proportionality.The enrichments were confirmed using DAVID 6.8 from which only enrichment clusters with P-value≥.01were considered.

Supplement Legends
Supplement Fig. S1A and B

.
Leukemia cell viability was measured by treatment with an 80-compound library targeting epigenetic protein families.The proliferative growth of 9 leukemia cell lines encompassing both acute myeloid and lymphoblastic leukemia diseases was measured.The effectiveness of each compound was measured at 1µM (left) and 10µM (right) concentrations for 48h.Purple hues (ratio to untreated >1.25) indicate proliferative growth advantage and red hues indicate diminished proliferative growth (< 0.75) and a ratio of 0.76 to 1.24 indicate no effect.The compound library was clustered and color coded according to the targeted function of the epigenetic protein family ie, histone deacetylase inhibitors (HDACi), histone methyltransferases (HTMi), bromodomain (BETi), DNA methylase (DNMTi), sirtuins (SIRTi), and other.Compound name, drug ID (G4, B5, etc.), and color code are listed in the columns right of the drug class.Each proliferative measurement was conducted in 6 measurements, which were averaged.Ratios were calculated as treated/untreated with the averaged O.D. 460nm measurement.Supplement Fig. S2.Stroma-dependent protection of primary AML depicted in three diagrams A. Percent apoptotic response of AML Mono and AML Cocu treated with HDACi (C11, C3, A8, B8, and H7) for 48 h at 10μM concentration.The hues represent the HDACi compound.B. The significance apoptotic difference of AML Mono versus AML Cocu as determined by paired student t-test (symbols *, **, and *** indicate P-value < 0.05, < 0.01, < 0.001, respectively).C. AML ID with unique hue colors to distinguish the AML samples.D. Percent apoptotic response of OCI-AML3 and K562 cell lines treated with HDACi (C11, C3, A8, B8, and H7) for 48 h at 10μM concentration.All values in a paired t-test between monoculture and coculture were significant (p-value <0.001) for HDACi tested.Supplement Fig. S3A left and right, A, Quantitation of phosphosites in ACSS2 S30 and ACACA S80 detected in both Mix1 and Mix2.Values are normalized and plotted as Log2 values.Light (L), Medium (M), and Heavy (H) refers to the isotope label.B, Western blot depicts representative protein levels of ACACA (Ser79) and -actin in mono and cocultured KG1a cells treated with HDACi, Api (5hr, 1M).Supplement Fig. S4A, CRISPR-editing of ACSS2 at S30 site (exon 1) with western blot of HS-5 single cell clones expanded of transfected cells with CRISPR plasmids targeting ACSS2.Representative lysates of HS-5 ACSS2-KO clones 1-6 show no ACSS2 expression compared to HS-5 WT whereas clone 7 depicts a clone not selected for further investigation.B, HS-5 ACSS2-KO single cell clones (Cl1-Cl6) proliferative growth compared to HS-5 WT.C, Apoptosis levels of KG1a cultured with HS-5 WT or with HS-5 ACSS2-KO (clones 1 and 2).The KG1a mono and cocultures (HS-5 WT or ACSS2-KO) were treated with HDACi (Api, SAHA, or CBHA).All concentrations were at 10μM.D, The FCCP-stimulated OCR was used to calculate spare respiratory capacity, the difference between maximal respiration (the maximal energy demand the leukemic cells can achieve) and basal respiration (the energy demand under basal conditions).Most of the basal respiration is coupled to ATP-linked respiration.These data complement the representative data shown in Fig. 4A.E, Leukemic proliferative growth with increasing acetate (0-50 M, 48h) dose is induced in KG1a, OCI-AML3, and K562.Supplement Fig. S5A, A representative flow cytometry histogram of KG1a cells treated with acetate (10M) or with Api (HDACi, 1M) 48h is shown.The intracellular expression of H3 and H3K9ac was assessed by flow cytometry.B, Positive cells were determined by a shift in fluorescence intensity relative to unstained cells.The percentage of positive cells were averaged for four replicates for H3 and H3K9ac measurements.Paired T-test was used to calculate the significance difference between untreated versus acetate or HDACi treatment.Shown is the significance of increased expression of H3 (p-value <0.003) and H3K9ac (p-value <0.01) with either acetate (10µM, 48h) or HDACi (Api, 1µM, 48h) treatment relative to untreated control.Supplement Table.S1.Clinical and mutational characteristics of AML samples used for proteomic profiling.Supplement Table.S2.TCGA groups A and B with the clinical, cytogenetic, and molecular variables are evaluated.A, Characteristics of the Groups (1 and 2) defined by the levels of expression of genes in the signature are shown in the Complex karyotype was defined by the presence of 3 or more cytogenetic aberrations.The covariate analysis reports the stratification of Group 2 with covariates, showing the p-value of Group 2 from the mutation (mt).B, The table depicts factors with a significant effect on overall survival probability for the TCGA cohort.