18F-FDG PET as an imaging biomarker for the response to FGFR-targeted therapy of cancer cells via FGFR-initiated mTOR/HK2 axis

Rationale: The overall clinical response to FGFR inhibitor (FGFRi) is far from satisfactory in cancer patients stratified by FGFR aberration, the current biomarker in clinical practice. A novel biomarker to evaluate the therapeutic response to FGFRi in a non-invasive and dynamic manner is thus greatly desired. Methods: Six FGFR-aberrant cancer cell lines were used, including four FGFRi-sensitive ones (NCI-H1581, NCI-H716, RT112 and Hep3B) and two FGFRi-resistant ones (primary for NCI-H2444 and acquired for NCI-H1581/AR). Cell viability and tumor xenograft growth analyses were performed to evaluate FGFRi sensitivities, accompanied by corresponding 18F-fluorodeoxyglucose (18F-FDG) uptake assay. mTOR/PLCγ/MEK-ERK signaling blockade by specific inhibitors or siRNAs was applied to determine the regulation mechanism. Results: FGFR inhibition decreased the in vitro accumulation of 18F-FDG only in four FGFRi-sensitive cell lines, but in neither of FGFRi-resistant ones. We then demonstrated that FGFRi-induced transcriptional downregulation of hexokinase 2 (HK2), a key factor of glucose metabolism and FDG trapping, via mTOR pathway leading to this decrease. Moreover, 18F-FDG PET imaging successfully differentiated the FGFRi-sensitive tumor xenografts from primary or acquired resistant ones by the tumor 18F-FDG accumulation change upon FGFRi treatment. Of note, both 18F-FDG tumor accumulation and HK2 expression could respond the administration/withdrawal of FGFRi in NCI-H1581 xenografts correspondingly. Conclusion: The novel association between the molecular mechanism (FGFR/mTOR/HK2 axis) and radiological phenotype (18F-FDG PET uptake) of FGFR-targeted therapy was demonstrated in multiple preclinical models. The adoption of 18F-FDG PET biomarker-based imaging strategy to assess response/resistance to FGFR inhibition may benefit treatment selection for cancer patients.


Introduction
Tyrosine kinase receptor fibroblast growth factor receptors (FGFRs) consist of 4 members including FGFR1, 2, 3, and 4, which play critical and diverse roles in early embryonic development and maintaining body metabolic balance. FGFR aberrant activation via gene fusion, activating mutation and amplification as well as ligand stimulation can promote tumor initiation and development in a variety of cancers [1][2][3]. For example, FGFR1 amplification was found in approximately 6% of lung cancer cases, mainly in squamous non-small cell lung carcinoma subtype without effective treatments [1,4]. FGFR2 was amplified in less than 10% of gastric cancer cases, associated with bad prognosis [1,2]. FGFR2 fusion occurred in 45% of intrahepatic cholangiocarcinoma cases [5]. Notably, fusion and activating mutations of FGFR3 frequently occurred in urothelial bladder carcinomas and predominantly in non-muscle invasive urothelial cell carcinoma type (occurring in 75% of cases) [2,6,7]. FGF19 amplification-induced FGFR4 activation was observed in hepatocellular carcinomas and might represent FGFR4-dependent cancer subtype [1,2,[8][9][10]. FGFRs thus become attractive targets for anti-cancer drug development. The pan-FGFR inhibitor (FGFRi) Erdafitinib, active against FGFR1-4, is the first FGFRi approved by FDA to treat patients with locally advanced or metastatic urothelial carcinoma with susceptible FGFR3 or FGFR2 genetic alterations, which has progressed during or following platinum-containing chemotherapy in 2019 [11,12]. Immediately afterward, FDA granted the approvals of FGFR1-3 inhibitors (Pemigatinib and Infigratinib) for patients with previously treated, unresectable locally advanced or metastatic cholangiocarcinoma with FGFR2 fusion or other rearrangements in 2020 [13] and in 2021 [14], respectively. Besides these three approved inhibitors, numerous FGFRi inhibitors, such as Rogaratininb, AZD4547, and Futibatinib, are still in phase I-III clinical trials in various malignancies [15,16].
Despite a promising prospect of FGFRi in certain cancer patients, its clinical efficacy is far from satisfactory with an overall response rate of 20-40% in the approved indication [17][18][19]. Even much lower response rate was observed in other cancer types [1,15,20]. Primary and acquired resistances to FGFR therapy due to the secondary mutations and the feedback activation of alternate pathways may decrease their clinical benefits [1,15]. It is indicated that the patients stratifying strategy based on FGFR aberration alone is very limited and cannot guarantee the patients' response to FGFRi. Exploring the therapeutic response biomarkers of FGFRi is therefore an urgent need. We have identified c-Myc, a fundamental downstream effector of FGFR signaling, could determine the therapeutic response to FGFRi in FGFR-addicted cancers [21]. Usually, c-Myc expression levels are examined in the tumor samples via biopsy or surgery, as well as in the circulating tumor cells (CTCs). However, sample availability, sensitivity on CTCs-based assays and tumor heterogenicity are unable to guarantee the accurate assessment of c-Myc. A novel biomarker to evaluate the therapeutic response to FGFRi in patients in a non-invasive, real-time and quantitative manner is greatly desired.
The identification of several FGFs, such as FGF1, FGF15/19, FGF21 and FGF23, with high relevance to metabolic regulation [22][23][24][25] attracted our attention. In fact, deregulating cellular metabolism, the hallmark of cancer, is required by the tumor cells to meet energy and structural requirements for rapid proliferation [26,27]. Notably, these metabolic changes are indispensable for certain cancers, making such tumors with metabolic vulnerability [26]; therefore, they can be exploited as therapeutic intervention or monitoring therapeutic response. We noticed that aberrant FGFR1 could enhance the Warburg Effect to drive prostate cancer progression by reprogramming LDH isoform expression and activity [28]. Meanwhile, the biologic basis for 18 F-fluorodeoxyglucose ( 18 F-FDG) positron emission tomography (PET) is the Warburg Effect [29]. 18 F-FDG is a radiolabeled analogue of glucose whereby the 2' hydroxyl group is substituted with 18 F. FDG passes the cellular membrane mediated by the glucose transporters (GLUTs) and is phosphorylated by hexokinases (HKs) to FDG-6-phosphate, which cannot be further metabolized to participate the tricarboxylic acid cycle [30]. Dephosphorylation of FDG-6-phosphate back to FDG by glucose-6phosphatase is the only way to exit the cells. The enhanced levels of GLUTs and HKs, along with the reduced level of glucose-6-phosphatase in tumors lead to FDG trapping in cancer cells [31]. 18 F-FDG PET imaging has thus been most widely applied in clinical practice and become the gold standard for oncology [32].
Considering the huge clinical translation potential, we investigated in this study whether 18 F-FDG PET could be used as a biomarker candidate for the therapeutic response to FGFRi in oncology. In different preclinical tumor models in vitro and in vivo, we found FGFR-targeted therapy decreased the accumulation of 18 F-FDG only in FGFRi-sensitive tumors, but not in FGFRi-resistant ones. We then demonstrated that downregulation of hexokinase 2 (HK2) via mTOR pathway by FGFR inhibition leading to the decrease of 18 F-FDG uptake in FGFRi-sensitive cells. A novel application of the well-established 18 F-FDG PET imaging to functional assessment of the treatment response to FGFR-targeted therapy in cancer patients as well as the underlying molecular mechanism are suggested.

Cell culture and reagents
NCI-H1581, NCI-H716, NCI-H2444 and Hep3B cells were obtained from the American Type Culture Collection (USA). RT112 cell was obtained from Deutsche Smmlung von Mikroorganismen und Zellkulturen GmbH (Germany). All cell lines in this study were maintained in the appropriate medium as suppliers suggested and were authenticated via short tandem repeats (STR) analysis with the latest test in 2020 (Genesky Biotechnologies, China) or single-nucleotide polymorphism (SNP) analysis with the latest test in 2021 (Crown Bioscience, China). All cells were routinely tested for mycoplasma by Mycoplasma Detection Kit-QuickTest (B39032; Biotool, China) and found to be free of contamination.
FGFR inhibitors (Erdafitinib, AZD4547, and BLU9931), AKT inhibitor MK2206, mTOR inhibitor AZD8055, proteasome inhibitor MG132 and lysosome inhibitor Leupeptin were purchased from Selleck Chemicals (China) and dissolved in DMSO at the concentration of 10 mM as a stock solution for in vitro study. AZD4547 was dissolved in 1% Tween-80 and BLU9931 was formulated in 0.5% carboxymethylcellulose/1% Tween-80 respectively for in vivo study.
To generate NCI-H1581 cells with acquired resistance to FGFRi, NCI-H1581 cells were treated by AZD4547 with the increasing concentration in a stepwise manner (from 30 nM to 1 μM). After approximate 6 months of induction, the NCI-H1581/AR cell line was obtained till its growth kinetics was similar to that of the parental NCI-H1581 cell line [33].

Mass spectrum
Protein extraction, digestion, Tandem Mass Tag (TMT) labeling and high pH reversed-phase liquid chromatography peptides fractionation were performed as described previously [34]. Detailed methods were available in the Supplementary Materials. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [35] partner repository with the dataset identifier PXD032227.

F-FDG uptake in vitro
For adherent cells, 2.0×10 5 cells/well were seeded in 12-well plates and 12 h FGFRi treatment at indicated concentrations in normal medium was started next day, followed by another 12 h FGFRi treatment in the glucose-free starvation medium with 5% fetal bovine serum. For non-adherent cells, 3.0×10 5 cells/well were seeded in 12-well plates in normal medium with FGFRi at indicated concentrations for 12 h, followed by another 12 h FGFRi treatment in the above starvation medium. After starvation, 1 µCi 18 F-FDG/well was then added and incubated at 37°C for 1 h. The radioactivity from both cell-accumulated and free 18 F-FDG were measured by an automatic gamma counter (Wizard 2470; PerkinElmer, USA). Relative 18 F-FDG uptake rate was normalized by cell number analyzed using Countstar BioTech (China). All samples were tested in triplicate.

Cell viability assay
Cells were inoculated in 96-well plates overnight and incubated with FGFRi at indicated concentrations or vehicle (as a negative control) for 72 h. Cell Counting Kit-8 (Dojindo Molecular Technologies, China) was used to assess cell viability as the instruction described. The normalized cell viability (%) was calculated as 100 × (ODFGFRi/ODvehicle).

Gene silencing by siRNA
Cells were inoculated in 6-well plates overnight and transfected with siRNAs as below by Oligofectamine RNAiMAX reagent (Invitrogen, USA) according to the manufacturer's instructions. After 48 h, cells were harvested for further analysis.

Animal studies
All animal studies were approved by the Institutional Animal Care & Use Committee of Shanghai Institute of Materia Medica, Chinese Academy of Sciences. 4-to 6-week-old female athymic nude mice nu/nu or SCID mice were provided by Shanghai Institute of Materia Medica or purchased from Beijing HFK Bioscience (China). 1×10 7 tumor cells, including NCI-H1581, Hep3B, NCI-H2444, or NCI-H1581/AR cells, were suspended in 200 μl ice-cold sterile PBS and subcutaneously injected into right flank of the mouse. Tumor-bearing mice were divided into the vehicle group and FGFRi treatment group randomly when the tumor volume reached approximately 100 mm 3 . For NCI-H1581, NCI-H1581/AR, and NCI-H2444 xenograft-bearing mice, AZD4547 (12.5 mg/kg, p.o., once a day) was given for 4 or 5 days. For Hep3B xenograft-bearing mice, BLU9931 (30 mg/kg, p.o., twice a day) was given for 6 days. Tumor size was measured by caliper every day and tumor volume (TV) was calculated with the formula: TV = (width 2 × length) / 2. The relative tumor volume was normalized by the TV immediately before FGFRi treatment.

PET/CT imaging
The tumor-bearing mice were fasted for 8 h before injection of 100-200 µCi 18 F-FDG via tail vein. During the uptake period (40-60 min), the mice were anesthetized under 1.5% isoflurane. Ten-min static data of PET imaging were recorded, followed by 10-min CT scan, using a microPET/CT scanner (Inveon; Siemens, Germany). PET data were reconstructed using the microQ Viewer software (Version 1.7.0.6; Siemens). Region of interest (ROI) delineating the tumor was drawn and Mean Standardized Uptake Value (SUVmean) of the tumor was obtained for 18 F-FDG uptake in vivo.

Statistical analysis
Data were presented as mean ± SD, except the data for xenograft growth curve, which were presented as mean ± SEM. The differences between two groups were analyzed by an unpaired Student's t-test using GraphPad Prism 8.0 software (USA). FGFR4-HK2 signature score was determined by Cox model [36]. Briefly, this score of each patient was calculated as follows: FGFR4-HK2 signature score = XFGFR4βFGFR4 + X HK2 β HK2 (X and β indicated the mRNA level and the risk coefficient of Cox model by survival analysis in R version 3.5.3, respectively). The correlations between the levels of FGFR4 mRNA/HK2 mRNA/FGFR4-HK2 signature score and overall survival (OS) were analyzed by the Kaplan-Meier method. The differences in the survival rates between curves were assessed by the log-rank test. p < 0.05 was considered statistically significant.

FGFR inhibition led to 18 F-FDG uptake reduction in the FGFRi-sensitive cancer cells
To identify novel biomarkers for evaluating the therapeutic response to FGFRi, TMT-labeled mass spectrometry-based proteomics was carried out in a FGFRi-sensitive cell line (NCI-H1581, a lung cancer cell line with FGFR1 amplification) upon AZD4547 treatment ( Figure 1A). As expected, proteins associated with cell cycle regulation were identified among the significantly differentially expressed ones (fold change > 1.2 or < 0.8, with p < 0.05; Table S1), which was consistent with our previous report [21]. Notably, protein levels of HK2 and GLUT3/14 (alias SLC2A3/14), the key factors highly related to glucose metabolism, especially to FDG trapping, were significantly decreased in AZD4547-treated NCI-H1581 cells than those in vehicle-treated ones ( Figure 1A). Therefore, 18 F-FDG uptake was tested in NCI-H1581 cells with FGFRi treatment in vitro. Two selective inhibitors, AZD4547 targeting FGFR1-3 and Erdafitinib targeting FGFR1-4 were used. We observed that both AZD4547 (0.1 μM) and Erdafitinib (0.01 μM) not only inhibited the cell proliferation (p < 0.001) but also reduced the 18 F-FDG uptake (p < 0.05) ( Figure 1B).
Whether FGFR inhibition leading to 18 F-FDG uptake reduction is a common effect on FGFR-aberrant tumor cells was further investigated. Four other cancer cell lines, the FGFRi-sensitive ones including NCI-H716 colon cancer cell line with FGFR2 amplification, RT112 bladder cancer cell line with FGFR3 amplification, and Hep3B liver cancer cell line with FGF19 amplification-induced FGFR4 activation, along with the FGFRi-primary resistant one (NCI-H2444 lung cancer cell line with FGFR1 amplification), were examined. Another FGFR4 selective inhibitor, BLU9931 was chosen for tests in Hep3B cells. Consistently, FGFRi-induced significant cell proliferation inhibition was still accompanied with 18 F-FDG uptake decrease in NCI-H716 cells (p < 0.001; Figure 1C), RT112 cells (p < 0.01; Figure 1D), and Hep3B cells (p < 0.01; Figure 1E). However, in NCI-H2444 cells both AZD4547 and Erdafitinib did not show the inhibitory effects on cell proliferation and 18 F-FDG uptake even at the concentration of 1 μM ( Figure 1F). These data implied that 18 F-FDG uptake might be correlated with the drug sensitivity/ resistance to FGFRi in FGFR-aberrant cancer cells.

FGFR inhibition downregulated HK2 gene via mTOR pathway
Since mass spectrometry-based proteomics identified that AZD4547 could decrease the protein levels of HK2 and GLUT3/14 ( Figure 1A), which are the main mediators of 18 F-FDG uptake, we tested the expression levels of these two molecules and other members in HK and GLUT families by Western blot analysis in NCI-H1581 cells for confirmation ( Figure  2A). Due to the subtle change of GLUT3 and the specific expression of GLUT14 major in testis [37], only the significant inhibitory effects of FGFRi on HK2 expression levels were further investigated. Three FGFR inhibitors (Erdafitinib, AZD4547 or BLU9931) were tested correspondingly in five FGFR-aberrant cancer cells. In the FGFRi-sensitive cells, including NCI-H1581 ( Figure 2B), NCI-H716 ( Figure 2C), RT112 ( Figure 2D), and Hep3B cells ( Figure 2E), FGFRi reduced the phosphorylated levels of FGFR (p-FGFR) or FRS2 (p-FRS2), which is the FGFR key adaptor protein as the well-recognized surrogate for FGFR activation [10,11,21,38], as well as decreased the protein levels of HK2. But in the FGFRi-resistant cells (NCI-H2444 cells), even p-FRS2 was suppressed by AZD4547 or Erdafitinib treatment, HK2 protein did not show significant changes ( Figure  2F). It was suggested that FGFRi could inhibit HK2 only in the FGFRi-sensitive cells.
How FGFR inhibition downregulated HK2 was then studied. Considering the FGFR aberration usually activates AKT-mTOR, PLCγ, and MEK-ERK pathways in cancer [1,2], we used the selective inhibitors or specific siRNAs to block these downstream signalings to test which could downregulate HK2. As Figure 2G-H shown, both the AKT inhibitor (MK2206) and the mTOR inhibitor (AZD8055) reduced the HK2 levels and 18 F-FDG uptake, as same as the FGFRi. However, neither knockdown PLCγ by siRNAs in NCI-H1581 cells ( Figure 2I) nor MEK inhibition by PD0325901 in NCI-H1581 and NCI-H716 cells ( Figure 2J) affected the HK2 levels. We also knockdown c-Myc, which was the downstream effector of FGFR via MEK-ERK signaling in FGFR aberrant cancer [21], and no obvious HK2 expression change was exhibited in NCI-H1581 and Hep3B cells ( Figure S1), indicating the different regulatory mechanisms by FGF/FGFR for HK2 and c-Myc. Herein, function of FGFRi's tumor inhibition may not be actioned simply by FGFR pathway; FGFR inhibition induced HK2 reduction via AKT-mTOR signaling to regulate glucose metabolism was indicated in the FGFRi-sensitive cells.
Whether FGFR inhibition downregulated HK2 expression at posttranslational level was next addressed. Neither the proteasome inhibitor MG132 nor the lysosome inhibitor Leupeptin could reverse HK2 downregulation induced by FGFR inhibition in NCI-H1581 ( Figure 2K), NCI-H716 ( Figure 2K) and Hep3B cells ( Figure  S2), suggesting this downregulation was not greatly dependent on protein degradation. mRNA levels of HK2 gene in NCI-H1581, NCI-H716, and NCI-H2444 cells were then detected at different time points (0-18 h after FGFRi treatment). (E) and NCI-H2444 (F) cells were incubated with indicated FGFR inhibitors (AZD4547, Erdafitinib, or BLU9931) at different concentrations. Cell viability (upper panels) and 18 F-FDG uptake (lower panels) were examined after 72 h and 24 h, respectively. Cells treated with vehicle were used as the normalization controls. Relative 18 F-FDG uptake was normalized by cell number. Data were shown as mean ± SD. *, p < 0.05; **, p < 0.01; ***, p < 0.001 vs vehicle group, using Student's t-test.
The significant decrease of HK2 mRNA was observed in FGFRi-sensitive cells starting from ~6 h after FGFR inhibition ( Figure 2L). Accordingly, HK2 protein level exhibited a slight reduction beginning from 6 h after FGFRi treatment and achieved a significant decrease after 12-or 24-h treatment in those cells ( Figure 2M). In FGFRi-resistant NCI-H2444 cells, no significant change of HK2 mRNA was detected under the treatment of AZD4547 or Erdafitinib ( Figure 2L). Furthermore, mTOR inhibitor (AZD8055) decreased expressional levels of HK2 mRNA ( Figure S3A) and HK2 protein in both FGFRi-sensitive ( Figure 2G) and -resistant cells ( Figure S3B). However, the mTOR signaling ( Figure  S3C) and HK2 protein ( Figure 2F) did not show significant changes by AZD4547 or Erdafitinib treatment in NCI-H2444 cells, implying that FGFR might lose its regulation on mTOR signaling, but mTOR could still modulate the expression of HK2 gene in this FGFR-resistant cell line.
The correlation between FGFR/HK2 signaling and liver cancer patients' prognosis was tested in TCGA-LIHC (liver hepatocellular carcinoma) dataset. The patients with OS information (n = 373) were classified into two groups based on the levels of FGFR4 mRNA, HK2 mRNA and FGFR4-HK2 signature score, respectively (median value as the cutoff). The Kaplan-Meier survival analysis showed that high levels of FGFR4 mRNA (p = 0.0232), HK2 mRNA (p = 0.0057) and FGFR4-HK2 signature score (p = 0.0027) were associated with poor OS of LIHC patients ( Figure S4). The importance of FGFR/HK2 signaling in the prognosis of LIHC patients was indicated.
Taken together, above results suggested that FGFR inhibition downregulated HK2 gene transcription via AKT-mTOR signaling, leading to the decrease of glucose uptake in the FGFRi-sensitive tumor cells.

F-FDG PET as an imaging biomarker for the therapeutic response to FGFRi in vivo
In order to investigate whether 18 F-FDG PET could be used as an imaging biomarker for the therapeutic response to FGFRi in vivo, the FGFRi-sensitive xenografts, NCI-H1581 ( Figure 3A) and Hep3B ( Figure 3B), as well as the FGFRi-primary resistant xenografts (NCI-H2444; Figure 3C) were generated for FGFRi treatment with visualization by 18 F-FDG PET/CT imaging in vivo. NCI-H1581 and NCI-H2444 xenograft-bearing mice were treated with AZD4547 (12.5 mg/kg, daily) for 5 days; and Hep3B xenograft-bearing mice were treated with BLU9931 (30 mg/kg, twice a day) for 6 days. Upon FGFRi treatment, SUVmean for 18 F-FDG probe was significantly reduced in NCI-H1581 xenografts (p < 0.01; Figure 3D) and in Hep3B xenografts (p < 0.001; Figure 3E); meanwhile, vehicle could not induce such decrease. In NCI-H2444 xenografts, marked change of 18 F-FDG probe was not detected in both AZD4547and vehicle-treatment groups ( Figure 3F). The following IHC analysis confirmed that FGFRi treatment led to the reductions of HK2 and Ki67 (a cell proliferation marker) only in the FGFRi-sensitive xenografts ( Figure 3G-H), but not in the FGFRi-resistant xenografts ( Figure 3I).
These data encouraged us to explore whether 18 F-FDG uptake in the FGFRi-acquired resistant tumors would be similar to that in the FGFRi-primary resistant tumors. NCI-H1581/AR cell line, which was previously generated from NCI-H1581 parental cell line by exposure to AZD4547 at concentrations increasing stepwise [33], was tested in vitro and in vivo. Its resistance to AZD4547 and Erdafitinib was validated ( Figure 4A). Consistent with the results from NCI-H2444 cells, 18 F-FDG uptake ( Figure 4B) and HK2 protein level ( Figure 4C) in NCI-H1581/AR cells did not show the remarkable changes in presence of AZD4547 or Erdafitinib. NCI-H1581/AR subcutaneous xenograft model was then created. Using the same dosage of AZD4547 as that in NCI-H1581 xenograft-bearing mice for 4 days, tumor growth could not be inhibited in NCI-H1581/AR xenografts ( Figure 4D). 18 F-FDG-based PET/CT imaging could not detect significant alterations of the probe accumulation in NCI-H1581/AR xenografts, responding to AZD4547 treatment ( Figure 4E). FGFRi treatment in vivo could not alter the protein levels of HK2 and Ki67 in NCI-H1581/AR xenografts ( Figure  4F). It was implied that the acquired resistance to targeted FGFR therapy might accompany by the disability of FGFRi-induced FDG uptake reduction.

Application of 18 F-FDG PET imaging to monitor the therapeutic response to FGFRi in vivo dynamically
To mimic the clinical practice of targeted FGFR therapy, we generated a NCI-H1581 xenograft model with the treatment regimen as a 5-day FGFRi treatment followed by a 4-day interval. 18 F-FDG PET/CT imaging was applied to assess the glucose uptake at three time points: 1) right before FGFRi treatment (Day 1); 2) right after FGFRi treatment (Day 6); 3) at the endpoint (Day 10; 4 days without FGFRi treatment) ( Figure 5A). Using this model, we wanted to investigate whether 18 F-FDG PET could reflect the tumor response to FGFRi in a dynamic and quantitative manner. NCI-H716 cells (lower panel) with AZD4547 treatment. β-Actin was used as the loading control. Data were shown as mean ± SD. **, p < 0.01; ***, p < 0.001; ns, p ≥ 0.05. Relative band intensity of target protein was normalized to its corresponding loading control as fold of the vehicle-, or control-, or non-treated group.
The tumor growth curves measured by calipers ( Figure 5B) showed that in the vehicle group, tumors kept growing with ~13.6-fold increase in tumor volume; in the AZD4547 group, tumor growth was significantly inhibited within the first phase (5-day with AZD4547), whereas tumor volume was slightly increased with ~1.9-fold increase in the second phase (4-day without AZD4547). As shown in Figure 5C, 18 F-FDG PET images demonstrated that vehicle treatment did not induced obvious changes in 18 F-FDG uptake by the tumor; and that 5-day FGFRi treatment resulted in a marked decrease of 18 F-FDG uptake by the tumor (p < 0.05). Notably, 4-day FGFRi withdrawal led to more 18 F-FDG accumulation in tumor (~2.3-fold increase in SUVmean on Day 10 compared with that on Day 6; p < 0.05), as well as the increased HK2 ( Figure 5D) and Ki67 levels ( Figure  5E). 18 F-FDG PET enabled the assessment of FGFRi therapeutic efficacy dynamically in vivo was indicated.

Discussion
FGFRs are clinically validated anticancer targets with pan-tumor potential, especially in the tumors lacking effective treatments. However, the clinical benefit in cancer patients with FGFR alterations is quite limited [39][40][41]. Moreover, even FGFR-aberrant patients attain an optimal response at an early stage, tumor relapse occurs eventually due to acquired resistance by the activation of bypass and downstream signalings or the development of FGFR secondary mutations [1]. Biomarkers or strategies with immediate translational potentials to evaluate the therapeutic response and to monitor the acquired resistance to FGFR-targeted therapy are urgently needed, particularly in a noninvasive and dynamic manner.
In the present study, TMT-labeled mass spectrometry-based proteomics suggested that FGFR inhibition regulated glucose metabolism in a FGFRi-sensitive cancer cell line NCI-H1581 with FGFR1 amplification. Interestingly, our previous report showed that cancer cells with FGFR-aberrant activation per se exhibited high glucose consumption into glycolytic pathway and resultant lactate production [42]. The critical role of glucose metabolism in FGFR-aberrant cancers is indicated, no matter with FGFRi treatment or not. In TCGA lung adenocarcinoma database (n = 740), upregulation of some glycolytic enzymes including HK2 gene was reported in FGFR-amplificated cancers, comparing with diploid cancers [42]. These findings encouraged us to further investigate whether FGFR kinasetargeted therapy was able to regulate HK2 expression and thereby inhibit glycolysis herein. Mechanistically, for the first time, we revealed that FGFR inhibition suppressed HK2 gene transcription via inhibiting mTOR in FGFRi-sensitive in vitro and in vivo cancer models with different FGFR1-4 anomalies ( Figure 6). In the FGFRi-sensitive cells, we found only mTOR inhibition could suppress HK2 expression; while inhibitions of other FGFR key downstream molecules (PLCγ and MEK/ERK) did not show the same inhibitory effect on HK2. Considering that c-Myc functioned as a key downstream effector in aberrantly activated FGFR signaling in cancer [21] and that FGF-induced vascular development was dependent on endothelial glycolysis via MYC/HK2 [43], we also tested whether HK2 downregulation by FGFRi was c-Myc dependent in this study. However, we found c-Myc knockdown had nonsignificant influence on HK2 expression level. Additionally, since HK2 can be regulated by several factors including epigenetic factors [44], perhaps HK2 expression might serve as a biomarker independently of FGFR aberration. Such different regulatory mechanisms on HK2 may partially be owing to the differences in biological context and cell lineage. The novel FGFR/mTOR/HK2 axis-mediated glucose metabolic regulation to assess the response to FGFR-targeted therapy is suggested. Nude mice bearing NCI-H1581 xenograft were orally treated with AZD4547 at 12.5 mg/kg once a day for 5 days, then AZD4547 was withdrawn for 4 days. 18 F-FDG PET/CT imaging was performed on Day 1 (before treatment), Day 6 (right after treatment) and Day 10 (after treatment interval). B, Relative tumor growth curves of NCI-H1581 xenografts (n = 3 for each group) within the indicated periods. Blue arrows, time points with 18 F-FDG PET/CT imaging; red arrows, time points with AZD4547 treatment.
Xenograft volumes at the starting point (0 day) were used as the normalization controls. Data were shown as mean ± SEM. C, 18 F-FDG PET/CT imaging on the NCI-H1581 xenograft-bearing animals upon FGFRi or vehicle treatment. Representative 18 F-FDG PET/CT images for SUVmean (left panel; xenografts were indicated with white dashed lines) and values for SUVmean from the xenografts (right panel) at the indicated time points were shown. Data were shown as mean ± SD. ns, p ≥ 0.05; *, p < 0.05. D-E, IHC staining for HK2 (D) and Ki67 (E) of NCI-H1581 xenografts on Day 6 (right after treatment) and Day 10 (after treatment interval). Scale bar, 50 μm. The disturbed glucose metabolism, as metabolic vulnerability for FGFR-addicted cancers, especially the corresponding expressional change of HK2, the key rate-limiting glycolytic enzyme, allows us trying to monitor the FGFRi-response by 18 F-FDG (an analog of glucose) PET imaging. We demonstrated that FGFR inhibition reduced 18 F-FDG in vitro uptake in four FGFRi-sensitive cancer cells, but not in primary and acquired resistant cancer cell lines to FGFRi. 18 F-FDG PET/CT imaging and IHC analysis on the tumor xenograft-bearing animals further confirmed the in vivo decreases of HK2 expression, 18 F-FDG tumor accumulation and tumor proliferation upon FGFRi treatment only in FGFRi-sensitive tumors; whereas these decreases were not observed in the FGFRi-de novo and acquired resistant tumors. The change of 18 F-FDG tumor uptake at early treatment stage might be used to identify the primary FGFRi-resistant patients noninvasively, despite the presence of FGFR activating mutations, to avoid unnecessary ineffective treatment and spare costs. Furthermore, both 18 F-FDG tumor accumulation and HK2 expression could respond the administration/withdrawal of FGFRi in NCI-H1581 xenografts correspondingly, which in turn suggested that the loss of 18 F-FDG tumor uptake response to FGFRi treatment might be associated with the acquired FGFRi-resistance. Certainly, using an inducible HK2 xenograft model monitored by 18 F-FDG PET imaging will further strengthen our novel finding, the association of FGFRi-regulated HK2 with 18 F-FDG uptake. 18 F-FDG PET as a novel biomarker for the response/resistance to FGFR-targeted therapy in cancers is thus indicated.
Since 18 F-FDG PET/CT imaging can provide both metabolic and anatomical information, it has been widely used in diagnosis, staging, molecular stratification and monitoring of the therapeutic effects and prognostic evaluation of cancer patients [45,46]. 18 F-FDG PET/CT imaging has been reported to monitor the therapeutic response to tyrosine kinase inhibitors (TKIs), such as EGFR- [47,48], VEGFR- [49], and ALK-TKIs [50]. The most successful application example of 18 F-FDG PET/CT imaging is prediction the survival outcomes and guidance the targeted therapy in thousands of non-small cell lung cancer with EGFR mutations involving hundreds of research articles [51][52][53]. However, no paper on 18 F-FDG PET/CT imaging for the drug sensitivity/resistance of FGFR-TKIs is published till now. Further literature search revealed that one related report on 18 F-FDG PET imaging was used to determine whether Dovitinib (a multitarget-tyrosine kinase inhibitor targeting FGFRs 1-3, VEGFRs, FLT3, c-Kit, PDGFR, and other receptor tyrosine kinases) altered tumor glucose metabolism and subsequent clinical outcome in a phase II study of 15 patients with recurrent or metastatic adenoid cystic carcinoma. 18 FDG-PET scans detected an early metabolic response only in 3 of 15 patients, but it did not correlate with RECIST (Response Evaluation Criteria in Solid Tumors) response. The authors claimed that they could not determine whether the observed effects were due to the specific inhibition of FGFR or other target receptors, or a combinatorial effect, because the enrolled patients were not selected for FGFR aberrance and Dovitinib was a multitarget kinase inhibitor [54]. Selective FGFR inhibitors in the selected patients with right drug target would be required to determine whether 18 FDG-PET could respond to the FGFR signaling inhibition in this rare cancer type.