Mitochondrial Raf1 Regulates Glutamine Catabolism

Abstract

Raf kinases play vital roles in normal mitogenic signaling and cancer, however, the identities of functionally important Raf-proximal proteins throughout the cell are not fully known. Raf1 proximity proteomics/BioID in Raf1-dependent cancer cells unexpectedly identified Raf1-adjacent proteins known to reside in the mitochondrial matrix. Inner-mitochondrial localization of Raf1 was confirmed by mitochondrial purification and super-resolution microscopy. Inside mitochondria, Raf1 associated with glutaminase (GLS) in diverse human cancers and enabled glutaminolysis, an important source of biosynthetic precursors in cancer. These impacts required Raf1 kinase activity and were independent of canonical MAP kinase pathway signaling. Kinase-dead mitochondrial matrix-localized Raf1 impaired glutaminolysis and tumorigenesis in vivo. These data indicate that Raf1 localizes inside mitochondria where it interacts with GLS to engage glutamine catabolism and support tumorigenesis.

One-Sentence Summary:

Raf1 is present within the mitochondrial matrix, where it binds GLS to regulate glutamine catabolism and tumorigenesis.

In cancer, Raf1 activation occurs via mechanisms that include mutation of upstream regulators, such as receptor tyrosine kinases and Ras GTPases, as well as by mutations that affect RAF1 itself, including via gene amplification (1–4). Once recruited to the plasma membrane (PM) Raf1 can engage downstream mitogen-activated protein kinase (MAPK) pathway signaling through phosphorylation of the MEK kinases (5). In addition to Raf1, A-Raf and B-Raf can also activate MEK and these other two Raf isoforms can compensate for MAPK activation in the event of Raf1 loss (6, 7). Despite this, Raf1 remains essential for the development and maintenance of some tumors through mechanisms independent of MAPK activity (7, 8). In this regard, Raf1 has well-described interactions outside the canonical MAPK pathway, including several with outer mitochondrial membrane (OMM) proteins (9, 10), although Raf1 has not been previously identified inside mitochondria. Mitochondria comprise a hub for various metabolic processes modulated in cancer cells to accommodate rapid proliferation. One such process is glutaminolysis, which involves the catabolism of glutamine to generate both ATP as well as precursors for the synthesis of fatty acids, nucleotides, and nonessential amino acids (11–13). Glutaminase (GLS) proteins, which catalyze the first and rate-limiting step of this process by converting glutamine to glutamate, are often upregulated in cancer (14–16). GLS activation has been previously associated with tumors driven by Ras, upstream regulators of Raf kinases (13, 17). Here we identify Raf1 protein inside mitochondria where Raf1 associates with GLS in the mitochondrial matrix to enable glutamine catabolism and tumorigenic growth.

Results

Raf1 proximal proteins nominated by BioID

To identify Raf1-adjacent proteins of interest in living cancer cells, the BASU promiscuous biotin ligase was fused to Raf1, then full-length fusion protein was expressed at endogenous levels and biotin labeling activity verified (Fig. 1A, fig. S1A–D) (18, 19). Raf1-BASU, but not eGFP-BASU fusion control, labeled positive control MEK MAPK proteins, demonstrating that the Raf1 fusion retains proximity to known physiologic interactors (fig. S1B). Raf1-BASU was then expressed in Raf1-dependent MM485 (melanoma) and AsPC1 (pancreas) cancer cell lines as well as matched Raf1-independent CHL1 and BxPC3 cancer cell lines derived from the same tumor types (20), namely melanoma and pancreatic cancer, respectively (fig. S1E). Biotin labeling followed by streptavidin pull-down and LC-MS/MS was then performed (18, 19). This nominated a number of proteins as proximal to Raf1, including 82 previously-identified Raf1 interactors, including Mek1, Mek2, YWHAQ, YWHAZ, B-Raf, and SPRY4 (Fig. 1B, table S1) (21). These findings indicate that this BioID approach can identify biologically relevant Raf1-associated proteins.

Fig. 1. Raf1 proteomics reveals mitochondrial localization.

(A) Schematic of Raf1 BioID and localization workflow. (B) SAINT plots demonstrating all hits with a positive SAINT score for proteins enriched in the proximal proteome of Raf1 in dependent MM485 or AsPC1 cells compared to Raf1-independent CHL1 and BxPC3 and eGFP controls. Red are select known Raf1 interactors and cyan indicates proteins of interest involved in metabolic processes. Dashed line drawn at SAINT score of 0.8 (C) Combined Cell Component, Biological Process, and Molecular Function Gene Ontology Enrichment in SAINT ≥ 0.8 proteins. Benjamini-Hochberg adjusted pValues are used. (D) Network of Raf1 proximal proteins with SAINT ≥ 0.9 in both pancreatic and melanoma cell lines of interest. Edges between non-Raf1 proteins represent known interactions. Proteins with known mitochondrial localization labeled with blue. (E) Mito-tag Mitochondrial isolation of MM485 melanoma cells. WCL denotes whole cell lysate, Mito indicates mitochondrial fraction, and PK denotes proteinase K treatment to digest outer mitochondrial membrane (OMM) proteins. Lamin A/C is a nuclear (nuc) protein, p70S6 kinase is a cytoplasmic (cyto) protein, and TOMM20 spans the mitochondrial outer mitochondrial membrane (OMM) (F) STED microscopy demonstrating inner-mitochondrial Raf1 (yellow arrows). Image stack side-view (XZ and YZ) are cropped laterally and then mean-projected along X or Y. Inset displays a Y-projection of a particular mitochondria of interest in XZ.

This dataset also included a number of novel putative Raf1-proximal proteins. Most notably, gene ontology analysis of interactors associated with Raf1 in Raf1-dependent cells revealed enrichment for proteins associated with the mitochondrial matrix (Fig. 1C, table S2), a subcellular location where Raf1 had not previously been described. In Raf1-dependent cancer cell lines, 38% of the most highly enriched Raf1-proximal proteins are known to localized inside mitochondria, among which is GLS (Fig. 1D) (14); such an enrichment was not seen in Raf1-independent cancer lines. This finding suggested a previously undescribed localization for Raf1 inside mitochondria in certain contexts.

To explore this possibility, mitochondria were purified then treated with proteinase K to remove outer mitochondrial proteins. Digestion of the extra-mitochondrial domain of the TOMM20 trans-membrane protein verified removal of proteins on the OMM; proteins inside mitochondria were retained, as demonstrated by detection of the inner-mitochondrial portion of TOMM20 (22). The Raf1-BASU fusion was observed in Raf1-dependent MM485 melanoma cells, but not in Raf1-independent CHL1 melanoma cells (fig. S1F). To further verify the localization of endogenous Raf1, mitochondria were isolated using the mito-tag protocol (23). Purified mitochondria treated with proteinase K confirmed an inner-mitochondrial localization of the endogenous Raf1 protein (Fig. 1E, fig. S1G). For orthogonal validation of Raf1 localization inside mitochondria, super-resolution stimulated emission depletion (STED) microscopy was performed using optical sectioning allowed by z-STED (Fig. 1F), confirming Raf1 localization inside mitochondria.

Raf1 impacts on glutaminase and MAPK pathway signaling

In addition to GLS, PCK2 and SUCLG2 were also among the mitochondrial proteins found to be proximal to Raf1 by BioID. The known functions of these mitochondrial matrix-localized proteins suggested that mitochondria-localized Raf1 might influence glutamine catabolism and the citric acid cycle (TCA). To explore this, ¹³C-glutamine tracing was performed with and without Raf1 depletion (Fig. 2A); MM485 was compared to CHL1 melanoma cells because BioID identified a stronger Raf1 mitochondrial localization in the former. Because GLS is the first and rate-limiting step of glutaminolysis (24), GLS depletion was also performed in parallel (fig. S2A), as a positive control to help benchmark the point at which glutaminolysis might be influenced by Raf1 activity. After glutamine starvation, labeled glutamine was reintroduced and metabolites were measured before steady state metabolite labeling occurred. Consistent labeling of glutamine was observed in both cell lines across all conditions, suggesting that glutamine uptake was unaffected by Raf1 loss in this setting (Fig. 2B, table S3). Loss of Raf1 reduced the labeling fraction (M+5) of glutamine-derived-glutamate as well as alpha-ketoglutarate in Raf1-dependent MM485 cells that contained Raf1 localized within mitochondria (Fig. 2B). This was also true for 2-hydroxyglutarate, which is derived from alpha-ketoglutarate (fig. S2C) (25). Further, labeled citrate demonstrated that Raf1 loss alters glutamine that is being processed reductively (M+5) as well as oxidatively (M+4) (Fig. 2C) (26). Raf1 has been previously linked to glutathione-S transferase P1 (GSTP1), however, no significant changes in the amount of labeled reduced glutathione were observed with Raf1 loss in MM485 cells (fig. S2D) (27). These data suggest that mitochondria-localized Raf1 may modulate the TCA at the level of GLS.

Fig. 2. Mitochondrial Raf1 regulates glutaminolysis.

(A) Schematic indicating ¹³C glutamine tracing for both reductive and oxidative TCA. (B) 13C-Glutamine tracing of M+5 Glutamine (Gln), Glutamate (Glu), and αKetoglutarate (αKG) with siRNA knockdown of GLS or Raf1; n = 3, **** P < 0.0001, ** P < 0.01. (C) Mass isotopologues of citrate pool in MM485 cells; n = 3,****P < 0.0001. (D) Glutamine to glutamate conversion as measured by luciferase-based Glutamine Glo assay with a non-targeting siRNA as a negative control. All Raf1 knockdowns rescued with differentially localized proteins. siCntrl (n =6) is nontargeting siRNA, siGLS (n=6), siRaf1 (n=6), and rescue with PM Raf1 (n=3), Mito Raf1 (n=6), or Mito Raf1 K375A kinase dead construct (n=3) ; **** P < 0.0001. (E) Western blot of MM485 cells blotting for ERK activation, effective Raf1 knockdown, Raf1 rescue, and successful GLS knockdown.(F) Quantitation of 2 western blots with phosphoERK normalized to total ERK.

To study if Raf1 localization inside mitochondria may be necessary for the observed effects on glutaminolysis, differentially-localized Raf1 constructs were produced. A mitochondrial matrix-localized Raf1 (Mito Raf1) was generated by fusing the protein to the mitochondrial-localized domain of COX4I1. As a control, a plasma membrane-localized Raf1 (PM Raf1) was created by adding the membrane-localization domain of the XRP2 protein to Raf1. Previous work demonstrated that PM Raf1 induces MAPK pathway activity (28). To assess Raf1 impacts on glutaminase activity, the conversion of glutamine to glutamate (29) was quantified. Congruent with glutamine tracing data, the amount of glutamine converted to glutamate decreased with Raf1 loss (Fig. 2D). Although it failed to activate the MAPK pathway, Mito Raf1 rescued the loss glutaminase activity, however, PM Raf1 did not (Fig. 2D), even though PM Raf1 induced the MAPK pathway (Fig. 2E–F). Mitochondrial delivery of the Raf1 K375A mutant, which alters Raf1 structure and abolishes Raf1 kinase activity (28), produced lower glutaminase activity than loss of Raf1 alone, suggesting a possible dominant-negative effect (Fig. 2D). These findings suggest that Raf1 effects on glutaminase activity and MAPK signaling are separable, and that mitochondrial-localized Raf1 mediates the former.

Raf1 association with GLS

The observations that Raf1 modulates glutaminase activity, that it can be detected in the mitochondrial matrix where GLS is canonically localized (14), and that it is proximal to GLS by BioID raised the possibility that Raf1 might interact physically with GLS. Consistent with Raf1 proximity to GLS, proximity ligation assay (PLA) in MM485 cells detected a Raf1-GLS signal; this signal was specific as it was diminished by Raf1 and GLS knockdown (Fig. 3A, fig. S3A–C). Additionally, Raf1 and GLS also displayed reciprocal co-immunoprecipitation; this was selective in that neither Ras nor MEK proteins brought down GLS (Fig. 3B, fig. S3D). Consistent with a mitochondrial location for Raf1-GLS interactions, Mito Raf1 robustly immunoprecipitated GLS whereas PM Raf1 did not (Fig. 3C). To examine the possibility that Raf1 and GLS proteins bind each other directly, purified recombinant proteins were studied by microscale thermophoresis (MST) (fig. S3E). Raf1 displayed a binding affinity to GLS (K_d=1.83×10⁻⁷M) comparable to its affinity for Mek1 (K_d=1.96×10⁻⁷M) (Fig. 3D). In contrast, another mitochondrial protein detected by Raf1 BioID, namely PCK2, failed to bind Raf1 (Fig. 3D), suggesting that associations between PCK2 and Raf1 in the mitochondrial matrix may be indirect. These data suggest that Raf1 can bind GLS directly at affinities comparable to well-characterized Raf1-interacting proteins.

Fig. 3. Raf1 directly interacts with GLS.

(A) PLA between Raf1 and GLS with shRNA knockdown of Raf1 with control. Scale bar represents 20 μm. Includes quantitation in bar graphs quantify particles per nucleus; *** P < 0.001 (B) Co-immunoprecipitation of Flag-6XHis-HA tagged MAPK component proteins with a flag antibody with appropriate inputs. (C) co-IP of plasma membrane and mitochondrial Raf1 with immunoblot for GLS with appropriate inputs. (D) Microscale thermophoresis of labeled 6X His-Raf1 against GLS, MEK1, or PCK2 to produce affinity constants of 183 nM, 196 nM, and no binding respectively. (E) Crosslinking mass spectrometry circos plot between crosslinked GLS and Raf1. Pink indicates the interface between Raf1 and GLS according to docking experiments. (F) Molecular docking between Raf1 in blue and GLS in green. Interfacing amino acids indicated in salmon.

To gain additional insight into the Raf1 association with GLS, interacting regions of Raf1 and GLS proteins were mapped. First, crosslinking mass spectrometry (CLMS) was used to identify points of interaction between Raf1 and GLS. Purified recombinant Raf1 and GLS protein were crosslinked to one another using bis(sulfosuccinimidyl)suberate (BS3) (fig. S3F) then mass spectrometry was performed. A number of crosslinked peptides were identified between Raf1 and GLS, with prominent signal between the N-terminal half of GLS and specific C- and N-terminal portions Raf1 (Fig. 3E, fig. S3G). Molecular docking simulations agreed with CLMS data (Fig. 3F), further supporting the existence of multiple Raf1-GLS contact interfaces in those protein regions.

Mitochondrial Raf1 in experimental tumorigenesis and spontaneous human cancers

The impact of mitochondria-localized Raf1 was next studied in experimental tumorigenesis. Wild-type Raf, PM Raf1, Mito Raf1, and Mito Raf1 K375A kinase were expressed in MM485 melanoma cells (fig. S4A). After subcutaneous injection in immune deficient mice, cells with enforced expression of wild-type and mito Raf1 displayed similar tumorigenic growth in vivo that were increased over PM Raf1 (Fig. 4A). In contrast, Mito Raf1 K375A expressing tumors grew significantly more slowly than others (Fig. 4A). Notably, Mito Raf1 K375A did not decrease ERK phosphorylation compared to cells not expressing the construct, suggesting that this abrogation of growth was not due to negative effects on global MAPK activity (Fig. 4B–C). Differences in tumorigenic growth in vivo were not reflected in proliferation in vitro (fig. S4B), indicating these impacts were not due to global impacts that alter cell viability or capacity for growth. To assess if the Raf1-GLS interaction occurs in spontaneous human malignancies, PLA was performed on a series of 26 spontaneous epidermal squamous cell carcinomas (SCC) (30), which are associated with Ras-MAPK activation, along with 6 independent normal skin controls. Epithelial cells in SCC displayed substantially more Raf1-GLS signal compared to normal epidermis on a per cell basis (Fig. 4D–E, fig. S5A–B). Raf1-GLS PLA on 42 additional human cancer specimens from breast, bladder, ovary, liver, pancreas, and prostate detected a range of Raf1-GLS PLA signals (fig. S5D–E). Taken together, these findings indicate that mitochondrial Raf1 can influence experimental tumorigenesis and that Raf1-GLS adjacency can be detected in a subset of spontaneous human tumors.

Fig. 4. Mitochondrial Raf1 and GLS interaction contributes to tumorigenesis and is present in patient tumors.

(A) Subcutaneous tumor growth with overexpression of 4 Raf1 constructs: WT Raf1 (n = 14), PM Raf1 (n = 8), Mito Raf1 (n = 8), and Mito Raf1 Kinase Dead (K375A) (n = 8); * P < 0.05, **** P < 0.0001. (B) Western blot of phosphorylated ERK and total ERK protein in cells expressing empty vector or mitochondrial Raf1 K375A. (C) With quantitation; P = ns. (D) PLA of spontaneous tumor microarrays for 3 normal and 3 squamous cell carcinoma (SCC) samples, including one matched. Scale bar represents 50 μm. (E) Density plot of puncta count per cell for PLA across several patient samples: 6 normal skin samples, and 26 SCC samples. Mean puncta count per cell is 6.928 for normal skin and 14.266 for SCC; **** P < 0.0001.

Discussion

These studies identified Raf1 inside mitochondria in certain contexts where it is proximal to a variety of proteins native to the mitochondrial matrix, including GLS. The mechanisms responsible for Raf1 translocation into mitochondria remain to be explored. Previous work, however, demonstrated HSP90-dependent mitochondrial translocation of other kinases, such as Akt1 (31). Mitochondria-localized Raf1 enables glutaminase activity and supports tumorigenesis without substantial impacts on MAPK signaling. This points to a non-canonical role for Raf1 in supporting glutaminolysis, a process which contributes biosynthetic precursors and energy in neoplasia. Impaired glutamine catabolism seen with global Raf1 loss can be rescued by mitochondrial-localized Raf1 but not Raf1 targeted to the plasma membrane, highlighting the potential role that Raf1 subcellular localization plays in this setting. The K375A Raf1 point mutation, which alters Raf1 protein conformation and kinase activity, disrupts both glutamine catabolism and tumorigenesis, underscoring the importance of the intactness of this Raf1 region in these effects. We were unable, however, to obtain evidence that GLS is a direct phosphorylation target of Raf1, providing a rationale for future efforts to define how Raf1 may modulate GLS activity. Recombinant Raf1 and GLS proteins associate directly at an affinity comparable to Raf1 binding to its well-studied downstream target, Mek1. CLMS and structure modeling nominated specific interface regions between GLS and Raf1. Evidence for Raf1-GLS proximity was observed in tumors from variety of tissues, suggesting that this interaction may play a potential role in spontaneous human cancers.