Pharmacologically targeting the myristoylation of the scaffold protein FRS2α inhibits FGF/FGFR-mediated oncogenic signaling and tumor progression

Fibroblast growth factor (FGF)/FGF receptor (FGFR) signaling facilitates tumor initiation and progression. Although currently approved inhibitors of FGFR kinase have shown therapeutic benefit in clinical trials, overexpression or mutations of FGFRs eventually confer drug resistance and thereby abrogate the desired activity of kinase inhibitors in many cancer types. In this study, we report that loss of myristoylation of fibroblast growth factor receptor substrate 2 (FRS2α), a scaffold protein essential for FGFR signaling, inhibits FGF/FGFR-mediated oncogenic signaling and FGF10-induced tumorigenesis. Moreover, a previously synthesized myristoyl-CoA analog, B13, which targets the activity of N-myristoyltransferases, suppressed FRS2α myristoylation and decreased the phosphorylation with mild alteration of FRS2α localization at the cell membrane. B13 inhibited oncogenic signaling induced by WT FGFRs or their drug-resistant mutants (FGFRsDRM). B13 alone or in combination with an FGFR inhibitor suppressed FGF-induced WT FGFR- or FGFRDRM-initiated phosphoinositide 3-kinase (PI3K) activity or MAPK signaling, inducing cell cycle arrest and thereby inhibiting cell proliferation and migration in several cancer cell types. Finally, B13 significantly inhibited the growth of xenograft tumors without pathological toxicity to the liver, kidney, or lung in vivo. In summary, our study suggests a possible therapeutic approach for inhibiting FGF/FGFR-mediated cancer progression and drug-resistant FGF/FGFR mutants.


Fibroblast growth factor (FGF)/FGF receptor (FGFR) signaling facilitates tumor initiation and progression. Although currently approved inhibitors of FGFR kinase have shown therapeutic benefit in clinical trials, overexpression or mutations of
FGFRs eventually confer drug resistance and thereby abrogate the desired activity of kinase inhibitors in many cancer types. In this study, we report that loss of myristoylation of fibroblast growth factor receptor substrate 2 (FRS2␣), a scaffold protein essential for FGFR signaling, inhibits FGF/FGFR-mediated oncogenic signaling and FGF10-induced tumorigenesis. Moreover, a previously synthesized myristoyl-CoA analog, B13, which targets the activity of N-myristoyltransferases, suppressed FRS2␣ myristoylation and decreased the phosphorylation with mild alteration of FRS2␣ localization at the cell membrane. B13 inhibited oncogenic signaling induced by WT FGFRs or their drug-resistant mutants (FGFRs DRM ). B13 alone or in combination with an FGFR inhibitor suppressed FGF-induced WT FGFR-or FGFR DRM -initiated phosphoinositide 3-kinase (PI3K) activity or MAPK signaling, inducing cell cycle arrest and thereby inhibiting cell proliferation and migration in several cancer cell types. Finally, B13 significantly inhibited the growth of xenograft tumors without pathological toxicity to the liver, kidney, or lung in vivo. In summary, our study suggests a possible therapeutic approach for inhibiting FGF/FGFR-mediated cancer progression and drug-resistant FGF/FGFR mutants.
Fibroblast growth factor (FGF) 2 /FGF receptor (FGFR) signaling regulates the fundamental development of multiple organs (1). However, a large body of research has demonstrated that this signaling axis is highly deregulated in numerous cancers, including amplification of FGFs in epithelial or stromal cells and/or aberrant expression or activation of FGFRs resulting from genetic translocation, mutation, or amplification in tumorigenic cells (2). Pathological FGF/FGFR signaling promotes cross-talk of tumorigenic cells with their microenvironment, which drives tumor proliferation, angiogenesis, and metastasis in cancer progression (3)(4)(5)(6)(7)(8).
FGF/FGFR signaling requires recruiting a scaffold protein called the fibroblast growth factor receptor substrate 2 (FRS2) to initiate downstream signaling. The FRS2 family is composed of two members, FRS2␣ and FRS2␤. Both proteins contain the phosphotyrosine-binding domain and multiple tyrosine phosphorylation sites (9,10). FRS2␣ mainly associates with FGF/ FGFR signaling. It binds to the juxtamembrane region of FGFRs through its phosphotyrosine-binding domain (10, 11), and the activation of FGFRs phosphorylates several tyrosine sites of FRS2␣. Whereas four sites mediate the binding with Grb2, which activates phosphatidylinositol 3-kinase/AKT signaling and to a lesser extent Ras/ERK signaling, the other two sites facilitate binding with SHP2, which activates mainly the Ras/ ERK pathway (10). Therefore, FRS2␣ is essential for FGF/ FGFR-induced signaling and facilitates cancer cell proliferation and migration (12). The aberrant expression of FRS2␣ is also observed in some cancers (13). Therefore, targeting FRS2␣ is considered an important therapeutic approach in the inhibition of FGF/FGFR-mediated tumorigenesis (14).
Myristoylation of FRS2␣ is essential for its anchoring to the plasmatic membrane. FRS2␣ contains the MGXXX(S/T) consensus sequence at the N terminus for N-myristoylation modification (10). N-Myristoyltransferase (NMT) catalyzes the myristoylation modification process by transferring the myristoyl group from myristoyl-CoA to the glycine at the N terminus of a protein (15). In this study, we investigated a therapeutic approach to inhibit FGF/FGFR-mediated oncogenic signaling and proliferation of cancer cells by blocking myristoylation of FRS2␣. We demonstrate that genetic knockout of FRS2␣ expression inhibited FGF/FGFR-mediated signaling. Signaling was rescued by the re-expression of the WT FRS2␣, but not FRS2␣(G2A), a mutation that prevents FRS2␣ myristoylation and FGF10-induced tumorigenesis. Additionally, a previously synthesized myristoyl-CoA analog inhibitor (16), B13, inhibited FRS2␣ myristoylation, resulting in reduced FGF/FGFRmediated oncogenic signaling and suppression of cell proliferation and migration of cancer cells by causing cell cycle arrest. The compound suppressed oncogenic signaling mediated by FGFR drug-resistant mutants and inhibited growth of xenograft tumors without observed pathological toxicity to host organs in vivo. Our study provides a potential therapeutic approach for the treatment of FGF/FGFR-mediated tumor progression.

Myristoylation of FRS2␣ is required for FGF2-induced signaling
Upon FGF induction, FRS2␣ is phosphorylated by FGFRs and further activates associated downstream signaling, such as PI3K/AKT and MAPK (9). To confirm the essential role of FRS2␣ in FGF/FGFR signaling, FRS2␣ was knocked out using the CRISPR/Cas9 system in mouse fibroblast cells NIH-3T3 (Fig. 1A). Two single colonies, FRS2␣⌬-1 and FRS2␣⌬-2, were identified and confirmed by PCR analysis (Fig. 1B) and Western blotting (Fig. 1C). Loss of FRS2␣ indeed impaired FGF2-induced AKT phosphorylation and MAPK signaling in both selected null mutants without alteration of PLC␥1 phosphorylation, which is not directly regulated by FRS2␣ activity (Fig.  1D). It should be noted that levels of p-ERK were much less inhibited than those of p-AKT due to the loss of FRS2␣. It has been reported that the activated PLC␥ induced by FGF/FGFR signaling could enhance MAPK signaling through the protein kinase C-Raf-MAPK pathway (17). As a result, although MAPK signaling is inhibited by knockout of FRS2␣, activation of PLC␥ could maintain MAPK signaling under the induction of FGF2.
The N terminus of FRS2␣ has been reported to be myristoylated (18). To examine whether myristoylation of FRS2␣ is crucial for FGF-induced signaling and its membrane localization, FRS2␣⌬ knockout cells were transduced with FRS2␣(WT) or FRS2␣(G2A) mutant by lentiviral infection (Fig. 1E). FRS2␣(WT), but not FRS2␣(G2A), a mutant unable to be myristoylated, rescued FGF2-induced AKT phosphorylation and MAPK signaling (Fig. 1F). Of note, levels of p-AKT and p-ERK were not able to be restored in the cells expressing FRS2␣(WT) (without FGF2 induction) (Fig. 1F, lane 5 versus lane 1). Higher expression levels of FRS2␣ tended to have a more inhibitory effect on p-ERK levels (Fig. S1). This is probably because the ectopic expression of FRS2␣ leads to negative feedback mediated by MAPK signaling, as reported previously (19).

Loss of FRS2␣ myristoylation inhibits FRS2␣ tyrosine phosphorylation and interaction with downstream proteins, SHP2 and GRB2
FRS2␣ contains numerous tyrosine and threonine phosphorylation sites (10). We examined whether loss of myristoylation leads to changes in phosphorylation status. Indeed, the levels of FRS2␣ tyrosine phosphorylation were significantly inhibited in cells expressing FRS2␣(G2A) compared with FRS2␣(WT) (Fig.  1G). In contrast, levels of FRS2␣ threonine phosphorylation were largely maintained in cells expressing FRS2␣(G2A) (Fig.  1G). Of note, FRS2␣(G2A) runs at a lower molecular weight in the gel compared with FRS2␣(WT) which is probably due to loss of tyrosine phosphorylation sites. In addition, loss of myristoylation resulted in reduced protein-protein interactions between FRS2␣ and GRB2 or SHP2 (Fig. 1H). This result suggests that overexpression of FRS2␣(G2A) is incapable of restoring FGF2-induced signaling in FRS2␣⌬ knockout cells (Fig. 1F) due to an interference with signal transduction.

A myristoyl-CoA analog, B13, suppresses the myristoylation of FRS2␣ and its cell membrane localization
Protein myristoylation is catalyzed by N-myristoyltransferases (NMT1 and NMT2) (22). Based on the NMT1 crystal structure (15), a myristoyl-CoA analog, B13, was identified from a panel of previously synthesized small-molecule compounds (23). An N-terminal truncated mutant of NMT1 was purified to examine the inhibitory effect of B13 (15). Compound B13 showed an inhibition of NMT1 enzymatic activity with IC 50 of 79.1 M (Fig. 3B). Of note, it has been reported that the NMT1 mutant lacking the N terminus shows an approximately 3-fold increase in enzymatic activity compared with the full-length enzyme (25). Therefore, the obtained IC 50 might lead to an underestimation of the inhibitory efficacy of B13 on protein myristoylation in the cell-based assay (Fig. 3C). B13 showed dose-dependent inhibition of FRS2␣ myristoylation detected by click chemistry (Fig. 3C). For the click chemistry Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors reaction, cells were incubated with myristic acid-azide (as a probe of labeling myristoyl-proteins) as the source of myristate for the myristoylation reaction. Cell lysates underwent a click reaction in which the azide functional group on myristoylated proteins was coupled to an alkyne tagged with biotin, allowing detection with streptavidin (Fig. 3C, lane 2 versus lane 1) (26). Myristoylation is essential for the localization of FRS2␣. FRS2␣ was mainly detected in the membrane fraction in FRS2␣(WT) cells, but mainly in the cytosol fraction in FRS2␣(G2A) cells (Fig. 3D).
FRS2␣ was predominantly expressed in the cell membrane fraction (Fig. 3E), probably because it has multiple acylation sites in the N terminus ( 2 GSCCSCP 8 ). FRS2␣ contains multiple potential S-palmitoylation sites (cysteines 4, 5, and 7) in addition to a myristoylation site (27). FRS2␣ levels increased about 2-fold in the cytosol fraction of the cells treated with B13 com-

Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors
pared with the control (Fig. 3E). Given the total amount of FRS2␣ in the cell membrane, loss of myristoylation only mildly affects membrane protein levels. However, consistent with the genetic results, B13 suppressed endogenous levels of p-FRS2␣ at the membrane and inhibited the phosphorylation of FRS2␣ under FGF2 induction (Fig. 3E). These data suggest that B13 regulates FRS2␣ activity and thus FGF/FGFR signaling.

B13 inhibits dysregulated FGFR oncogenic signaling in a variety of cancer cells
Similar to the genetic loss of FRS2␣ in NIH-3T3 cells (Fig. 1, D and F), B13-mediated inhibition of FRS2␣ myristoylation reduced the levels of p-FRS2␣, p-AKT, and p-ERK in a dose-dependent manner (Fig. 4A) and to a lesser extent p-PLC␥1 (Figs. 1D and 3G). Additionally, the inhibition required more than 12 h of treatment at a concentration of 15 M (Fig. 4B).
Dysregulation of FGF/FGFR signaling, including mutations resulting in FGFR activation and/or FGFR gene amplification, has been demonstrated to drive proliferation and survival of cancer cells. B13 reduced the levels of p-FRS2␣ and p-AKT with or without induction by FGF2 in NIH-3T3 cells ectopically expressing FGFR2 but did not affect the level of p-FGFR (Fig.  4C). These data suggest that B13 inhibits FGF-induced signaling by targeting FRS2␣ and bypasses the activation of FGFRs.
Dysregulated FGFR signaling due to overexpression or activation of FGFR1/2 has been reported in a variety of cancer types. Bioinformatics analysis on expression levels of FGFR1/2 in 947 cancer cell lines indicated that FGFR1 expression levels were elevated in numerous cancer cell lines, including MDA-MB-134-VI breast cancer cells, and FGFR2 expression levels were elevated in SNU16 and KATO III stomach cancer cells (Fig. S3, A and B). PC-3, 22Rv1, and other prostate cancer cells expressed unaltered levels of FGFR1/2 compared with other cancer cell lines (28) (Fig. S3, A and B). The expression levels of FGFR1/2 were confirmed in a panel of selected cancer cells, including MDA-MB-134-VI, SNU16, KATO III, AGS, HCT116, and prostate cancer cells LNCaP, 22Rv1, PC-3, DU145, and normal prostate cells PNT2 (Fig. S3C). FRS2␣ expression was also detected in MDA-MB-134-VI, SNU16, KATO III, AGS, and HCT116 cells, which are also recommended by ATCC for anti-FGFR drug discovery analysis (Fig. S3C).
We further examined the inhibitory effect of B13 on endogenous FGFR-mediated signaling in some of these cancer cell lines. The levels of p-FRS2␣, p-AKT, and/or p-ERK were elevated by FGF2 induction in MDA-MB-134-VI and SNU16 cells, but not KATO III, probably due to high expression of FGFR2 (Fig. S3C). B13 significantly inhibited the levels of pFRS2␣, p-AKT, and/or p-ERK under basal levels or exogenously added FGF2 in the examined cancer cells (Fig. 4, D-F), suggesting that the compound inhibits FGFR-mediated oncogenic signaling in a variety of cancer cells.

B13 collaborates with FGFR inhibitors to reduce FGF-induced signaling
Because B13 inhibits the FRS2␣ scaffolding function, we further examined whether B13 collaborated with FGFR inhibitors, such as PD173074 or dovitinib (a receptor tyrosine kinase inhibitor), which target FGFR1/2 activity. Because cancer cells were under FGF2 induction, the inhibitory effect was specifically due to FGFR-induced signaling. Whereas PD173074 or dovitinib alone reduced the levels of p-FRS2␣, p-AKT, and p-ERK in cells with endogenous (

B13 inhibits cancer cell proliferation and migration through cell cycle arrest
FGF/FGFR oncogenic signaling promotes cancer cell proliferation, survival, migration, invasion, and angiogenesis (6). Therefore, we examined whether proliferation of cancer cells via FGF/FGFR-mediated signaling depends on the expression of FRS2␣. Indeed, shRNA-FRS2␣ inhibited the proliferation of AGS and KATO III cancer cells (Fig. S6, A and B), both of which possess elevated FGF/FGFR oncogenic signaling. Next, we examined whether cell proliferation depended on the myristoylation of FRS2␣ by treating cells with B13. B13 significantly inhibited the proliferation of SNU16, KATO III, and AGS cells (Fig. 6, A-C); however, B13 or genetic knockdown of FRS2␣ had limited inhibition on HepG2 (liver cancer cells) or 293T cells (Fig. S7, A-D). SNU16 and AGS cancer cell lines were  The cell lysates were subjected to immunoprecipitation with anti-FLAG antibodies. Immunoprecipitated proteins underwent the click chemistry reaction with biotin-alkyne, and myristoylated proteins were detected using horseradish peroxidase-conjugated streptavidin via immunoblotting. D, genetic mutation leading to loss of myristoylation inhibits the association of FRS2␣ with the cell membrane. FRS2␣⌬ cells were transduced with control vector (V), FRS2␣(WT), or the myristoylation-deficient mutant FRS2␣(G2A), and the expression levels of FRS2␣ were detected in the cytosol and cell membrane fractions. E, the cell membrane protein fractions were isolated from NIH-3T3 cells treated with/without B13 (15 M) and with/without FGF2 induction (50 g/ml, 10 min). 100 g of the cytosol proteins (60 l) and 20 g of the cell membrane proteins (20 l) were loaded. The protein levels of total FRS2␣ and p-FRS2␣ were measured in each fraction. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and caveolin-1 were markers of the cytosol and cell membrane fractions, respectively. Immunoblots of FRS2␣ at long and short exposure times are displayed. Error bars, S.D.

Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors
particularly sensitive to B13 treatment with IC 50 values of 4.0 and 8.5 M, respectively. Additionally, B13 also inhibited cell migration of MDA-MB-134-VI (Fig. 6D). We further analyzed the role of B13 on the cell cycle and found that B13 significantly arrested the cell cycle of SNU16 cells at the G 2 /M phase, KATO III cells at the S phase, and AGS cells at the G 1 phase (Fig. 6,  E-G). The inhibition of the cell cycle was further confirmed by up-regulation of p21 and/or p27 expression or down-regulation

Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors
of CDK2, CDK4, and/or CDK6 (Fig. 6H). The cells arresting at different phases of the cell cycle by B13 and the alteration of different master regulator proteins might represent the diver-sity of cancer cell characteristics. Nevertheless, the myristoyl-CoA analog inhibitor suppressed proliferation of cancer cells by cell cycle arrest.

Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors B13 inhibits the growth of xenograft tumors in vivo
To evaluate the effect of B13 on the growth of xenograft tumors, we first examined which cancer cells might be more dependent on FRS2␣ expression for tumor growth in vivo. SNU16 (higher expression levels of FGFR2) and PC-3 cells (lower expression level of FGFR2) were chosen (Fig. S3C). These two cell lines expressing shRNA-control or shRNA-FRS2␣ were subcutaneously injected into SCID mice. The size and weight of SNU16 and PC-3 xenograft tumors expressing shRNA-FRS2␣ were significantly reduced compared with the control group (Fig. S8, A and B). IHC staining showed that expression levels of Ki-67 and CD34 in the FRS2␣ knockdown tumors were reduced compared with the control group (Fig. S8,  C and D). These data indicate that FRS2␣ is important for tumor growth in vivo. Additionally, the extent of inhibition of xenograft tumors by shRNA-FRS2␣ was higher in SNU16 xenograft tumors compared with PC-3 xenograft tumors (Fig.  S8, A and B), probably due to higher expression levels of FGFR2 in SNU16 cells, which is more dependent on FGF/FGFR-mediated oncogenic signaling (Fig. S3C).
After confirming the dependence on FRS2␣ for the growth of SNU16 xenograft tumors, we evaluated the inhibitory effect of B13. The inhibitor had no significant effect on the body weight of SCID mice (Fig. 7A) or histological structure of the liver, kidney, and lung (Fig. 7B). The size and weight of tumors were significantly reduced in the B13 group compared with the vehicle control (Fig. 7C). IHC staining showed that expression levels of Ki-67 and CD34 in tumors were lower in the B13-treated group compared with the control (Fig. 7D).
The results indicate that B13, the myristoyl-CoA analog inhibitor, has no observed toxicity to the major organs of the host mice but is effective for the treatment of cancer progression in a mouse model.

Discussion
Our study demonstrates a novel approach in targeting FGF/ FGFR-mediated oncogenic signaling and tumor progression. The co-translational myristoylation modification of FRS2␣, a scaffold protein of FGFRs, plays an essential role in regulating FGF/FGFR signaling. Genetic ablation of FRS2␣ myristoylation

Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors
suppresses FGF/FGFR-mediated AKT and/or MAPK activation (Fig. S9). Myristoylation promotes the association of FRS2␣ at the cell membrane, which might be required to facilitate the interaction of FRS2␣ with FGFRs. It is well documented that FGF/FGFR signaling facilitates the cross-talk of the epithelium with its microenvironment (9). For example, FRS2␣ has been illustrated as an important node in FGF/FGFR signaling in embryonic development (12). Additionally, FGF/ FGFR is also one of the oncogenic driver signaling pathways in numerous cancers (31). Therefore, targeting myristoylation will provide a therapeutic strategy in FGFR-mediated cancer (32).

Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors
eration and migration of a variety of cancer cells. Given the fact that the dysregulation of FGF/FGFR signaling (8,9) and amplification of FRS2␣ are associated with numerous high-grade cancer types (13,35,36), B13 will provide a therapeutic approach to inhibit FGF/FGFR-mediated tumor progression.
Targeting FRS2␣ myristoylation exhibits benefits over FGFR inhibitors in the suppression of FGF/FGFR-mediated tumorigenesis. Currently, numerous FGFR inhibitors, including PD173074, dovitinib, and ponatinib, that block the tyrosine kinase domain of FGFRs are undergoing clinical trials for cancer treatment (37)(38)(39). Although these drugs exhibit substantial clinical responses, nonsynonymous mutations have been identified among the FGFRs. A majority of tumors develop drug-resistant mutants with elevated FGFR activity (30, 40 -43). Among those, mutations of the gatekeeper residues, such as FGFR1(V561M) and FGFR3(V555M), have been shown to confer resistance to the multikinase inhibitor PP58 and the FGFR inhibitor AZ12908010, respectively (44). Because FRS2␣ is an immediate downstream node of FGFRs, the FRS2␣ myristoylation inhibitor will avoid a selection pressure on FGFRs but will exhibit a similar inhibitory effect on FGF/FGFR signaling. In particular, targeting FRS2␣ myristoylation will potentially bypass FGFR DRM -induced tumor progression. Additionally, our data indicate that the combination of a FRS2␣ myristoylation inhibitor together with FGFR-targeting drugs shows a synergistic effect for inhibiting FGF/FGFR signaling.
NMTs regulate the myristoylation of an array of oncogenic proteins. Many oncogenic proteins, including c-Src, AMPK (AMP-activated protein kinase), and MARCKS (myristoylated alanine-rich C kinase substrate) require myristoylation modification to carry out their cellular functions (45)(46)(47). Therefore, B13 might target tumor progression in which myristoylation is essential for the "addicted" oncogenic protein in the tumors (48). Whereas cancer cells, such as SNU16, KATO III, and MDA-MS-134-VI, possess aberrant expression of FGFR1/2, other cancer cells, such as PC-3, are dependent on Src kinase activity (26). B13 might serve as an agent for targeting tumor progression driven by the oncogenic pathways in which a myristoylated protein is essential to mediate the oncogenic signaling. Promisingly, our data show that normal cells are different from cancer cells in responding to B13 inhibition. Cancer cells usually up-regulate their fatty acid metabolism, which potentially provides an elevated amount of myristoyl-CoA to support protein myristoylation (49 -51). Additionally, B13 has also been reported as having an inhibitory effect on the proliferation of cancer cells via the regulation of ceramide biogenesis. The elevation of the ceramide to sphingosine ratio promotes the apoptotic pathway (52,53). Therefore, B13 might possess a dual effect by targeting myristoylation and ceramidase functions in cancer cells.
Numerous NMT inhibitors have been reported to inhibit fungal growth due to low amino acid sequence similarity of the NMT protein between fungi or parasites and humans (54, 55). However, only a limited number of agents have been identified so far as anti-cancer agents (15,22). B13 provides a possible therapeutic avenue to target oncogenic signaling pathways that require myristoylation, including FRS2␣-mediated FGF/FGFR tumor progression, and block the cross-talk of cancer cells with their microenvironment, leading to metastatic disease (8). Of note, FRS2␣ also mediates other receptor tyrosine kinases, such as vascular endothelial growth factor receptor, nerve growth factor receptor, and brain-derived neurotrophic factor receptor (56 -58), suggesting that B13 could potentially inhibit these receptor-mediated diseases as well.

Plasmid construction and lentiviral production
The coding sequence of mouse FGFR1, FGFR2 (FGFR2c isoform), or mutants was subcloned into the lentiviral vector FUCRW (59). The FRS2␣ mutant, FRS2␣(G2A) (loss of myristoylation site), was also cloned into FUCRW. The primer sequences used for cloning are listed in Table S1. The cloning details are described in the supporting Materials and Methods.
Lentivirus was generated by co-transfecting plasmids expressing the gene of interest and the packaging vectors MDL, VSV, and REV into HEK293T cells. Virus infection was performed as described previously (24). All lentivirus procedures followed the guidelines and regulations of the University of Georgia.

Cell culture
PNT2 cells were purchased from Sigma (catalog no. 95012613). NIH-3T3, 293T, and prostate cancer cell lines, including LNCaP, 22Rv1 DU145, and PC-3, were purchased from the American Type Culture Collection (ATCC) in September 2013. The FGFR genetic alteration cell panel, including SNU16, KATO III, AGS, HCT116, and MDA-MB-134-VI, was purchased from ATCC in October 2014. These cell lines carrying aberrant expression of FGFRs were recommended by ATCC for anti-FGFR drug discovery research. HepG2 was purchased from ATCC in March 2016. All of the cell lines purchased were defined as passage 1 (the first thawing) upon arrival. All cell lines were cultured in ATCC-recommended medium and temperature. Cell lines from ATCC and Sigma had a certificate of mycoplasma-free and authentication when purchased. All of the cell lines were used within no more than 8 -10 passages and were periodically examined for mycoplasma contamination by using the ATCC detection kit.

CRISPR/Cas9-mediated gene knockout
Plasmids for FRS2␣ knockout were purchased from Ori-Gene. One of the two guide RNA (gRNA) vectors (or scramble control vector) and donor vector (1:1 ratio) were co-transduced into NIH-3T3 cells by transient transfection using Lipofectamine 3000 (Thermo Fisher Scientific). After 48 h posttransfection, cells were split at a 1:10 ratio and cultured for an additional 3 days. After splitting cells seven times (at a 1:10 ratio), 2 g/ml puromycin was applied to select drug-resistant cells. The medium was changed every 3 days. Single colonies were isolated by the limiting dilution method. The genomic DNA and proteins were extracted from the single colonies. Genomic PCR was performed to check whether the fragment between the two homologous arms in the donor vector was integrated into the genome properly. The primer sequences used for genomic PCR are listed in Table S1. Additionally, the

Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors
genetic knockout of FRS2␣ in isolated colonies was confirmed by Western blot analysis.

Examination of myristoylation by immunoprecipitation and click chemistry
Cells were grown in 10-cm dishes to 80 -90% confluence. To metabolically label myristoylated proteins, cells were further cultured in medium containing 12-azidododecanoic acid (Thermo Fisher Scientific, C10268) and 2% BSA (fatty acidfree) for 24 h. Cells were washed twice with PBS, and then 1 ml of lysis buffer (50 mM Tris-HCl, pH 7.4, with 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100) containing protease inhibitors was added. Protein lysates were extracted by incubation with the lysis buffer for 30 min on a shaker and collected by scraping. The lysates were centrifuged for 10 min at 12,000 ϫ g. The supernatant was collected, and protein concentration was measured. Protein lysate (1 mg) was incubated with 40 l of prewashed anti-FLAG M2 affinity gel (Sigma-Aldrich, A2220) overnight. After immunoprecipitation, the resin was washed three times with lysis buffer and twice with M-PER buffer (Thermo Fisher Scientific). The Click-iT chemistry reaction was carried out for 1 h at room temperature in freshly made Click-iT reaction buffer (2 mM CuSO 4 , 1 mM bis(tert-butyl)tris(triazolylmethyl)amine-propanol, 10 mM sodium L-ascorbate, 50 M biotin-alkyne). The mixture was then suspended in 4ϫ SDS-loading buffer containing 2-mercaptoethanol. Samples were heated at 98°C for 5 min and separated on 10% Trisglycine SDS-polyacrylamide gels. Myristoylation was detected by probing the membrane with streptavidin-horseradish peroxidase (Thermo Fisher Scientific).

Subcellular fractionation
Cells from 10-cm dishes were lysed in subcellular fractionation buffer (250 mM sucrose, 20 mM HEPES (pH 7.4), 10 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 1 mM DTT, and protease inhibitor mixture). The lysates were passed through a 25-gauge needle 10 times using a 1-ml syringe and then left on ice for 20 min. The nuclear fraction was collected by centrifugation at 720 ϫ g for 5 min. The supernatant was transferred to a new tube and centrifuged at 10,000 ϫ g for 5 min to pellet the mitochondrial fraction. The new supernatant containing cytosolic and cell membrane fractions was centrifuged in an ultracentrifuge at 100,000 ϫ g for 1 h. The supernatant (cytosolic fraction) was transferred to a new tube. The pellet was washed with 400 l of the fractionation buffer and resuspended by pipetting. The sample was recentrifuged for 45 min. The cell membrane pellet was resuspended in lysis buffer with 10% glycerol and 0.1% SDS.

Fluorescence-based NMT activity assay
The assay was performed in 96-well black microplates (Greiner Bio-One, Germany). Each well contained 31.

Cell proliferation, migration assay, and cell cycle analysis
For the proliferation assay with drug treatment, cells were seeded in 96-well plates at a density of 3000 -5000 cells per well depending on the growth rate of each cell line. After overnight incubation, cell numbers were measured by the MTT assay as day 0. Cells were then treated with different B13 concentrations or DMSO (a control) with each treatment having five independent wells as repeats. Cell numbers were measured after 24 h of treatment. The remaining wells were replaced with fresh medium containing B13 or DMSO every 24 h for 5 consecutive days. For the MTT assay, the growth medium was replaced with 100 l of fresh culture medium without phenol red along with 10 l of 12 mM MTT (Thermo Fisher Scientific) stock solution (dissolved in PBS) or 100 l of the medium as a blank control. After incubation at 37°C for 4 h, 100 l of 10% SDS (dissolved in 0.01 M HCl) was added to each well. The plate was incubated at 37°C for 4 h in a humidified chamber. Each sample was mixed by pipetting, and the absorbance at 570 nm was measured.
Cells were transduced with control vector or shRNA by lentiviral infection, and proliferation was measured. The transduced cells were cultured for 3 days and then plated at the same density (day 0). The cell numbers were measured by the MTT assay for 5 consecutive days. Cancer cell migration was evaluated by the wound-healing or the Transwell assay as described in the supporting Materials and Methods.
For cell cycle analysis, cells were grown in a medium containing 50 ng/ml FGF2 with B13 or DMSO for 3 days. The culture medium was replaced with fresh medium (with/without compound and FGF2) every 24 h. Cells (1 ϫ 10 6 ) were collected and washed once with PBS and fixed with ice-cold 70% ethanol for Inhibiting myristoylation of FRS2␣ in FGFR-mediated tumors were stained by FxCycle TM PI/RNase solution (Thermo Fisher Scientific) for 30 min in the dark at room temperature. Cells were analyzed using flow cytometry (CyAn ADP) (488-nm excitation and 585-nm emission).

Xenograft tumors
To examine the inhibitory activity of B13 on the growth of xenograft tumors, 2 ϫ 10 6 SNU16 cells were subcutaneously inoculated in the flank side of SCID female mice. After 2 weeks, mice carrying xenografts were randomly separated into two groups. B13 was dissolved in the vehicle solution containing 30% Kolliphor, 65% saline (0.9% NaCl), and 5% ethanol. One group received 160 l of the drug solution at a concentration of 65 mg/kg body weight via tail vein injection twice a week for 6 weeks. The other group received 160 l of vehicle via tail vein injection twice a week for 6 weeks. Body weight and tumor size were measured (length ϫ width ϫ width/2) weekly. Xenograft tumors, liver, lungs, and kidneys were harvested for immunohistochemistry analysis.
For examining the effect of FRS2␣ knockdown on xenograft tumors, SNU16 cells and PC-3 cells were transduced with shRNA targeting FRS2␣ or control shRNA by lentiviral infection. The infected SNU16 cells (2 ϫ 10 6 ) or PC-3 cells (5 ϫ 10 5 ) were subcutaneously inoculated in the flank side of SCID female and male mice, respectively. After 8 weeks (SNU16) or 6 weeks (PC-3), xenograft tumors were harvested. All animals were maintained according to the surgical and experimental procedures of the protocol A2013 03-008 approved by the institutional animal use and care committee at the University of Georgia.

Statistical analysis
Prism software was used to carry out statistical analyses. The data are presented as mean Ϯ S.D. or mean Ϯ S.E. and analyzed using Student's t test. All t tests were performed at the twosided 0.05 level for significance. *, p Ͻ 0.05; **, p Ͻ 0.01.