HSF1 remodels mitochondrial biogenesis and function in cancer cells via TIMM17A

Abstract

Mitochondria play critical roles in energy production and cellular metabolism. Despite the Warburg effect, mitochondria are crucial for the survival and proliferation of cancer cells. Heat Shock Factor 1 (HSF1), a key transcription factor in the cellular heat shock response, promotes malignancy and metastasis when aberrantly activated. To understand the multifaceted roles of HSF1 in cancer, we performed a genome-wide CRISPR screen to identify epistatic interactors of HSF1 in cancer cell proliferation. The verified interactors of HSF1 include those involved in DNA replication and repair, transcriptional and post-transcriptional gene expression, and mitochondrial functions. Specifically, we found that HSF1 promotes cell proliferation, mitochondrial biogenesis, respiration, and ATP production in a manner dependent on TIMM17A, a subunit of the inner membrane translocase. HSF1 upregulates the steady-state level of the short-lived TIMM17A protein via its direct target genes, HSPD1 and HSPE1, which encode subunits of the mitochondrial chaperonin complex and are responsible for protein refolding once imported into the matrix. The HSF1-HSPD1/HSPE1-TIMM17A axis remodels the mitochondrial proteome to promote mitochondrial translation and energy production, thereby supporting robust cell proliferation. Our work reveals a mechanism by which mitochondria adjust protein uptake according to the folding capacity in the matrix by altering TIM complex composition.

Fig. 1. HSF1 interactors are involved in critical cellular processes.

(A) Schematic diagram of epistatic interactions by comparing cell proliferation defects caused by single and double knockout (KO) of HSF1 and another essential gene ‘A’. The ratio (%) of viable cells compared to that in the control is plotted in the histograms. The additive effect: ratio (double KO) = ratio (HSF1-KO) * ratio (Gene A-KO). (B) Experimental strategy: A population of PC3M wild-type cells (PC3M-WT) or HSF1-KO cells was infected with a pooled sgRNA library and collected at different time points (every 3 to 4 days) during negative selection. The abundance of sgRNA-encoding constructs was determined by deep sequencing and compared to the input samples at Day 0. Two clones of PC3M-WT and HSF1-KO cells were used in the paired screens. (C) CRISPR score calculation using RAD51 as an example. The relative abundance of sgRNA over time was fitted into an exponential decay model. The beta value, representing how quickly the RAD51-KO cells declined in the cell population, is calculated as the CRISPR score. RAD51 sgRNA decreased more rapidly in the PC3M-WT cells than in the HSF1-KO cells. Correspondingly, RAD51 exhibits a smaller CRISPR score in PC3M-WT cells than in HSF1-KO cells. Dots on the graph represent the weighted average abundance of the four sgRNAs targeting RAD51 at each time point, which are utilized for the beta value calculation. (D) Scatter plots illustrate the delta CRISPR scores (PC3M-WT minus HSF1-KO) for genes essential in either cell type. Blue and red dots represent the overlapped synthetic lethal and genetic buffering interactors of HSF1, respectively, in both paired screens. (E) Schematic diagram of the co-CRISPR assay. The sgRNA targeting the MAPT1 gene or the AAVS locus serves as the control for HSF1 sgRNA during the first infection. The sgRNA of an HSF1 interactor or a non-target control (NTC) was co-expressed with Cas9 during the second infection. (F & G) Histograms showing the number of viable cells measured by PrestoBlue assays on Day 7. The ratios to the control + NTC cell numbers (F) and the ratios to the corresponding NTC in either the control or HSF1-KO cells (G) are presented (mean ± SD, n=3). Dashed lines represent the calculated additive effects. Unpaired t-test: ns P >= 0.05; * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001. (H) Network of verified epistatic interactors of HSF1. The physical and functional interaction network was retrieved from the STRING database and grouped using k-means clustering (n=3). The inner circle color of each node represents the cluster to which the gene belongs. The halo color of each node indicates the type of genetic interaction (red: genetic buffering; blue: synthetic lethal), while the darkness reflects the confidence (verified in one of the co-CRISPR experiments or both). The color saturation of edges denotes the confidence score of a functional interaction between two interactors.

Fig. 2. HSF1 remodels the steady-state mitochondrial proteome.

(A) Scatter plots displaying the log2 fold change (FC) of protein (x axis) and mRNA (y axis) of HSF1 epistatic interactors following the knockout of HSF1 in PC3M cells. The epistatic interactors of HSF1 that altered protein levels by at least 1.25-fold (FDR: 0.05) are included. RNA-seq was performed in biological triplicate, and TMT-proteomics was performed in four biological replicates. The average fold changes (HSF1-KO vs. PC3M-WT) in mRNA and protein levels are plotted. (B) Histograms showing gene ontology (GO) analysis of differentially expressed proteins following the knockout of HSF1 in PC3M cells (1.5-fold, FDR: 0.05). (C&D) Histograms displaying the numbers of differentially expressed proteins following HSF1 knockout in PC3M cells across various mitochondrial compartments. Mitochondrial genes that significantly altered their protein (C) or mRNA (D) levels (1.5-fold, FDR: 0.05) were included. IMS: intermembrane space; MIM: mitochondrial inner membrane; MOM: mitochondrial outer membrane. (E&F) Heatmap showing log2 fold change (FC) of protein and mRNA of differentially expressed mitochondrial genes upon knockout of HSF1 in PC3M cells. 61 genes significantly changed both their mRNA and protein levels (1.5-fold, FDR: 0.05) (E). 99 genes significantly changed either their mRNA (F, upper panel) or protein levels (F, lower panel) but not both.

Fig. 3. HSF1 promotes mitochondrial biogenesis and function.

(A) Schematic diagram of deuterium oxide (D₂O) labeling. (B&C) Histograms showing the ratios of newly synthesized mitochondrial proteins (B) and cytosolic proteins (C) to the newly synthesized DNA, as detected through deuterium oxide (D₂O) labeling in PC3M-WT and HSF1-KO cells. Samples were collected hourly from 2 to 6 hours of labeling (n=3). The mean and SEM of the 5 time points are shown. Paired t-test: ns P>=0.05; * P<0.05. (D) Western blot analysis of newly synthesized cytosolic and mitochondrial proteins after one hour of HPG labeling in PC3M-WT and HSF1-KO cells. Upper panel: HPG-labeled proteins were biotinylated in a click reaction and detected using streptavidin-HRP and ECL. Lower panel: Western blot (WB) analysis of GAPDH, MPARS15, and Histone H3 as markers for cytosolic, mitochondrial, and nuclear proteins, respectively. (E&F) Basal oxygen consumption rate (OCR) (E) and cellular ATP levels (F) in PC3M-WT and HSF1-KO cells.

Fig. 4. HSF1 alters TIM complex composition.

(A) Heatmap showing log2 fold change (FC) of protein and mRNA of selected mitochondrial genes upon knockout of HSF1 in PC3M cells. The 23 mitochondrial genes that reduced protein levels at least 1.5-fold more than the changes in their mRNA levels are shown, along with the two additional subunits of the TIM23 core complex. The asterisks indicate the subunits of the TIM23 core complex. (B&C) Representative images (B) and the quantification (mean ± SD, n=3) (C) of western blot analysis on the TIM23 core complex in PC3M-WT and HSF1-KO cells. (D) Newly synthesized mitochondrial proteins after one hour of HPG labeling in the PC3M cells with TIMM17A knockout or treated with a non-target control (NTC). Western blot (WB) analysis of MRPS15 serves as a loading control for steady-state mitochondrial proteins. (E-G) Histograms displaying the number of viable cells (E), basal oxygen consumption rate (OCR) (F), and cellular ATP levels (G) in PC3M cells with single or double knockout of HSF1 and TIMM17A. After antibiotic selection, cells were seeded on Day 0. Viable cells were measured by PrestoBlue assays on Day 4. OCR and ATP were measured on Day 1.

Fig. 5. HSF1 regulates TIMM17A protein via the mitochondrial chaperonin complex.

(A) Heatmap of HSF-1 occupancy at CUT&Tag peaks in PC3M cells. Normalized HSF-1 reads were mapped to 100 bp bins, ±1000 bp from the peak summits. (B) The top DNA motif enriched at the center of HSF1 CUT&Tag peaks (±100 bp from the peak summits). (C) Scatter plots of mRNA log2 fold change (FC) at HSF1-bound genes upon HSF1 knockout in PC3M cells. The 97 genes associated with HSF1 CUT&Tag peaks within 2kb of the transcription start sites are shown. Differentially expressed (DE) genes (1.5-fold, FDR: 0.05) are indicated in blue (down-regulated) or red (up-regulated) color. The top 10 DE genes with the most significant fold change are labeled with gene names. (D) Genome browser view of HSF1 occupancy at the head-to-head HSPD1 and HSPE1 genes. (E) Protein levels of HSPD1 and HSPE1 determined by TMT proteomics in PC3M wild-type cells (PC3M-WT) and PC3M cells with HSF1 knockout (HSF1-KO). Unpaired t-test: **** P < 0.0001. (F) Histograms showing the number of viable cells upon HSPD1 siRNA treatment in PC3M cells. (G&H) Representative images (G) and quantification (mean ± SD, n=3) (C) of western blot analysis on HSPD1, HSPE1, and TIMM17A upon HSPD1 siRNA treatment. In each experiment, the tested proteins were first normalized to the ACTIN loading control, and the ratios relative to the protein level in the control siRNA (ctrl) were calculated and plotted into histograms. Unpaired t-test (against control siRNA): ns P >= 0.05; * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001.

Fig. 6. TIMM17A promotes mitochondrial translation.

(A&B) Representative images (A) and quantification (mean ± SD, n=3) (B) of western blot analysis for mitochondrial translation factors, GFM2 and TARS2, in PC3M wild-type cells (PC3M-WT) and PC3M cells with HSF1 knockout (HSF1-KO). The asterisk indicates a higher molecular band recognized by the TARS2 antibody. Unpaired t-test: * P < 0.05; *** P < 0.001. (C) Western blot analysis of TARS2 knockout (TARS2-KO) in PC3M wild-type cells (PC3M-WT) and PC3M cells with HSF1 knockout (HSF1-KO). NTC: non-target control. The asterisk indicates a likely non-specific band recognized by the TARS2 antibody. (D&E) Coessentiality rank and correlation coefficient for all genes with TIMM17A (D) and TIMM17B (E). The positions of mitochondrial translation factors (Mito Translation) in the coessentiality plots are indicated as red dots. (F&G) Representative images (F) and quantification (mean ± SD, n=3) (G) of western blot analysis in PC3M cells following TIMM17A knockout. The asterisk denotes a likely non-specific band recognized by the TARS2 antibody. Either MAPT1 (n=4) or AAVS (n=2) was employed as the negative control in the co-CRISPR assays along with a non-target control (NTC) sgRNA. In each experiment, the tested proteins were normalized to the GAPDH loading control, and the ratios relative to the protein level in NTC were calculated. Results (n=6) were pooled for quantification and plotted into histograms. Unpaired t-test: ns P >= 0.05; ** P < 0.01; *** P < 0.001. (H&I) Representative images (H) and quantification (mean ± SD) (I) of HPG labeling for newly synthesized proteins by mitochondrial translation. HPG labeling was performed in PC3M cells with TIMM17 knockout (TIMM17-KO) or the non-target control (NTC) for 45 min in L-methionine-free medium containing cycloheximide (CHX) to inhibit cytosolic translation. MAPT1 was used as the negative control in the co-CRISPR assay. Immunofluorescence of ATP5 was used as a mitochondrial marker, and DAPI was used to stain DNA. Scare bar: 20 μM. Unpaired t-test: **** P < 0.0001.

Fig. 7. TIMM17A protein level is coupled with HSF1 activity to promote robust cell proliferation.

(A) Western blot analysis of TIMM17A upon YME1L1 knockdown. PC3M cells with HSF1 knockout (HSF1-KO) were treated with YME1L1 siRNA or control (ctrl) siRNA. YME1L1 and TIMM17A protein levels were monitored over time with GAPDH as a loading control. The tested proteins were normalized to GAPDH, and the ratios relative to those in the control siRNA are labeled. (B&C) Histograms showing the numbers of viable cells measured by PrestoBlue assays on Day 5 after siRNA treatment. The wildtype PC3M cells (PC3M-WT) and HSF1-KO cells were treated with YME1L1 siRNA or control (ctrl) siRNA. The ratios to the cell number in PC3M-WT + control siRNA (B) and the ratios to the cell numbers in the corresponding control siRNA (either PC3M-WT or HSF1-KO) (C) are presented (mean ± SD, n=3). Dashed lines represent the calculated additive effects. Unpaired t-test: ** P < 0.01; **** P < 0.0001. (D) A proposed model for HSF1-mediated mitochondrial remodeling via regulating the TIM23 core complex composition.