Tumor Drug Concentration and Phosphoproteomic Profiles After Two Weeks of Treatment With Sunitinib in Patients with Newly Diagnosed Glioblastoma

Despite currently available therapies, the prognosis and outcome of patients with glioblastoma (GBM) are poor. Clinical trials evaluating tyrosine kinase inhibitors (TKI) for recurrent GBM, including TKIs with a broad kinase inhibition profile such as sunitinib and vandetanib, resulted in minimal improvement in progression-free survival and no overall survival benefit, despite their significant clinical benefit in other malignancies (1–4). TKIs inhibit intracellular signaling, that is, tyrosine kinases, that exert a variety of biological activities in GBM, including cell proliferation and migration (5, 6). Whether their disappointing efficacy in GBM is due to restricted drug delivery of TKIs or a result of differential biological characteristics compared with other malignancies is unknown, but local TKI concentration reached upon treatment in the tumor may be hampered by the blood–brain barrier (BBB; ref. 7). While TKIs inhibit their target kinases at low concentrations, multiple other kinases that are involved in (GBM) tumor growth may be inhibited at higher concentrations (8, 9).

In addition, it is anticipated that the biological activity of specific kinases driving the growth of GBM in an individual patient is one of the factors that will determine response to treatment with TKIs (5). Analyzing tyrosine-phosphorylated proteins by mass spectrometry (MS)-based tyrosine-phosphoproteomics can provide insight in the effect of TKIs on cellular signaling (10–13). Initial promise has been generated on the identification of ALK-, ROS-, and PDGFRα-mediated non—small cell lung cancer (NSCLC) subtypes by this high-throughput method (14).

Based on these considerations, we hypothesized that in an individual patient both the achievable local GBM TKI concentrations and the specific activity of GBM-related cellular kinases, driving its malignant behavior, will influence potential treatment benefit.

To address aforementioned issues regarding (in)efficacy of TKI treatment in GBM, seeking to gain more insight in their activity at the tumor level in patients, we performed a pilot study of preoperative treatment with sunitinib in newly-diagnosed patients. Sunitinib tumor concentrations and its in vitro efficacy in GBM-cell lines were determined to study to what extent clinically achieved tumor concentrations are sufficient for inhibition of tumor cell proliferation. In addition, the effects of sunitinib treatment on tyrosine kinase signaling activities were analyzed.

This interventional pilot study was performed in Amsterdam UMC, location VUmc, Cancer Center Amsterdam (Amsterdam, the Netherlands) between January 2015 and October 2016, with preoperative, off-label use of the TKI sunitinib.The primary objective was to determine sunitinib tumor and cerebrospinal fluid (CSF) concentrations in patients with newly-diagnosed GBM, who were treated for 10 to 14 days prior to surgical resection (Supplementary Fig. S1). Based on previous studies reporting tumor concentrations, we aimed to include 5 patients to obtain reliable estimates of sunitinib tumor concentrations (15, 16). Tumor and CSF concentrations were compared with plasma concentrations in these patients and the in vitro sensitivity of GBM tumor cell lines. Secondary objectives were to determine the feasibility of tyrosine phosphoproteomic profiling in these tumors and to evaluate the effect of sunitinib treatment on tyrosine kinase signaling activity by MS-based tyrosine phosphoproteomics. Phosphoproteomic profiles of sunitinib-treated patients were compared with those obtained in resection specimens of a control group of patients with newly-diagnosed GBM without preoperative treatment.

The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines and approved by the Institutional Review Board (medical ethics committee) of Amsterdam UMC, location VUmc. All treated patients provided written study-specific informed consent. The patients who participated as a control group for the secondary objectives signed a general informed consent form from the Department of Neurosurgery of Amsterdam UMC, location VUmc, for use of their resected tissue for experimental purposes. This trial is registered within the Netherlands Trial Register (www.trialregister.nl, NL4609) and via www.ClinicalTrials.gov, identifier NCT02239952.

Eligible patients for preoperative treatment were adults (>18 years) with a newly-diagnosed brain tumor on initial MRI, suggesting resectable GBM according to interpretation of an expert neuro-radiologist. Main inclusion and exclusion criteria are described in Supplementary Methods.

Patients awaiting surgery received 2 weeks of treatment with the oral TKI sunitinib (50 mg once daily) prior to the initial resection of the tumor (Supplementary Fig. S1). The last dose of sunitinib was administered in the morning on the day of surgery. Venous blood sampling, for pharmacokinetic measurements of sunitinib (trough levels), was performed on day 8, the day before surgery, and approximately 5 days after surgery. Tumor tissue samples (approximately 1 cm³) and at least 30 μL of CSF were obtained during the tumor resection.

Immediately after tumor resection, tumor tissue was placed in an aluminium tissue container, snap-frozen in liquid nitrogen, and stored at −80°C until further processing. Next, multiple 10-μm cryosections of frozen tissue were made for the lysis procedure as well as for hematoxylin and eosin (H&E) staining to determine tumor cell content. In general, approximately 50% of the tumor tissue was used for determination of tumor drug concentrations. From the remaining tissue, cryosections were made for MS-based phosphoproteomics as described before (17) and further detailed in Supplementary Methods. Minimal protein input was set at 5.0 mg per tumor.

Concentrations of sunitinib in tumor tissue, CSF, and plasma were determined using LC-MS/MS as described previously (16), with further details in Supplementary Methods.

Four GBM cell lines, U87, U251, T98G, and U138, were used to determine in vitro cell proliferation and sunitinib IC₅₀ concentrations. The HT29 colorectal cell line was included as a positive control for comparison with prior experiments and response to sunitinib was assessed by MTT assay as described previously (details in Supplementary Methods; refs. 18, 19).

Peptides were separated by an Ultimate 3000 nanoLC-MS/MS system (Dionex LC-Packings) coupled online to a Q HF Exactive mass spectrometer (Thermo Fisher Scientific) as previously described (20, 21). MS/MS spectra were searched against the Uniprot human reference proteome FASTA file (release March 2017, 42161 entries, no fragments) using MaxQuant1.5.4.1 (22). Downstream data analysis and identification of phosphopeptides, phosphosites, and proteins is further described in Supplementary Methods.

Ranking of the top 20 activated kinases in samples from control and sunitinib-treated patients and visualization of kinase-substrate networks of individual samples was performed using Integrative Inferred Kinase Activity (INKA) analysis as described before (12). Differentially activated kinases were identified and significance was determined with a Mann–Whitney U test. Identification of “outlier kinases” in individual samples based on INKA scores was done by application of Tukey rule, i.e., values differ from the median more than 1.5 times the interquartile range, and further filtering for an INKA score >10.

Posttranslational modifications (PTM) signature enrichment analysis (PTM-SEA) was performed as described elsewhere (ref. 23; for details, see Supplementary Methods).

The MS proteomics data are available and were provided to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD019038 (24).

Six patients with newly diagnosed GBM signed informed consent and started preoperative sunitinib treatment, of whom 5 were evaluable for the primary analysis. One patient was replaced upon withdrawal of informed consent after 1 week of treatment. Median duration of preoperative sunitinib treatment was 13 days (range 11–18; Table 1). Study treatment was well tolerated and no grade 3 to 4 adverse events were observed. Tumor resection was performed without perioperative complications in all patients. GBM diagnosis was histopathologically confirmed for all patients. Nine patients with GBM undergoing surgical resection were included as a control group (Supplementary Table S1).

Sunitinib concentrations in tumor tissue and plasma were determined in all 5 patients. CSF could be collected during surgery in 4 of these patients. Median sunitinib tumor concentration was 1.9 μmol/L (range 1.0–3.4), 10-fold higher than the median sunitinib plasma concentration on the day of surgery (0.194 μmol/L, range 0.105–0.479). In CSF the median concentration was 0.015 μmol/L (range 0.013–0.023; Table 1). No relation between sunitinib tumor accumulation and use of dexamethasone and anticonvulsants was found (Supplementary Table S1).

Sensitivity of GBM cell lines U87, U251, T98G, and U138, and the colorectal cell line HT29 as a positive control was tested for a concentration range from 0 to 20 μmol/L of sunitinib incubated for 72 hours. Exposure to sunitinib resulted in a heterogeneous response in the GBM cell lines, with a median IC₅₀ value of 5.4 μmol/L (range 3.0–8.5 μmol/L), shown in Fig. 1. The IC₅₀ concentrations for the GBM cell lines were significantly higher (2.8-fold) than the detected intratumoral sunitinib concentrations in patients as described above (P = 0.01). As a positive control, the colorectal cancer cell line was most sensitive with an IC₅₀ similar to prior experiments and other cell lines previously tested (25).

Phosphotyrosine (pTyr)-phosphoproteomics was successfully performed in 12 of 15 tumor samples, of which 11 were suitable for comparative analyses between sunitinib-pretreated and control samples. Three samples (2 control and 1 sunitinib-treated patient) were excluded for downstream (comparative) analysis because of much lower protein input and significantly less phosphopeptide identifications (P = 0.03; Supplementary Table S1A). One patient (SUN-wk1) who withdrew consent after 1 week of sunitinib treatment was analyzed for potential drug targets only. For the 12 tumor samples with 5-mg protein input 1,557 ± 329 (SD) phoshopeptides per sample were identified out of 6,767 ± 837 of the total number of peptides per sample (23% phosphopeptide enrichment). A median of 1,411 phosphosites were identified per sample with 91% phosphorylation on tyrosine residues, indicating succesful pTyr enrichment (Supplementary Table S1B). All identified and quantified phosphopeptides and sites can be found in Supplementary Tables S2 and S3. Unsupervised hierarchical clustering of identified phosphopeptides did not differentiate between tumor samples from treated versus untreated samples (Supplementary Fig. S2). Supervised clustering of the differentially detected tyrosine-phosphopeptides (cut-off P < 0.05 without additional filtering) revealed a clear separation between sunitinib-treated and control tissues (Fig. 2). Between the four sunitinib-treated and seven control tissues, 108 phosphopeptides were up- and 23 significantly downregulated (cut-off P < 0.05). Figure 3 shows a protein interaction network of these differentially regulated phosphopeptides (listed in Supplementary Table S4) mapped to proteins, showing an EGFR-centered signaling network with upregulation of multiple other kinases.

In general the top 20 active kinases that were identified by INKA analysis of individual GBM samples were mostly comparable between different GBM samples, including GSK3A and 3B, MAPK10, YES, FYN, and SRC (Fig. 4). In three treated and one untreated tumor sample MAPK10 and in one treated tumor sample EGFR stood out as the highest active kinase. For this tumor sample (SUN 04-B), an EGFR mutation was detected [EGFR (c.865G>A; p.A289T)] and potential EGFR amplification (Fig. 4). Comparative analysis of the kinase activities identified in both groups showed differential and upregulated activity of FGFR3 (P = 0.036) and EGFR (P = 0.047) in sunitinib-treated samples (Supplementary Fig. S3A). Activity of the known sunitinib targets KDR (VEGFR2), PDGFRA, and PDGFRB was low and not significantly different between the groups. Outlier analysis of kinase activity scores, as a potential strategy to identify drug targets in individual tumors, revealed nine outlier kinases in four tumors. These were MAPK10 (INKA score > 250 in CON-08 vs. median score < 50 for other 10 samples), PRKCB, NTRK2 and NTRK3, INSR and IGF1R, PDGFRA and FGFR1, and EPHB2 (Supplementary Fig. S3B).

Based on PTM-SEA analysis involving 558 PTM signatures, three strongly significantly enriched phosphosite signatures were identified in sunitinib-treated compared to control samples (23). Two of these were based on Phosphosite-Plus (PSP) EGFR and SRC kinase activity signatures and one was a PSP-derived EGF perturbation signature. Two moderately significant NetPath derived signatures corresponding to EGFR and Notch signaling were also found (Supplementary Fig. S4).

Failure to achieve therapeutic TKI drug levels in brain tumors may be an underlying reason for their relatively poor efficacy. We found that treatment of patients with GBM with the TKI sunitinib at the standard dose of 50 mg every day resulted in a significantly lower median tumor concentration (1.9 μmol/L) than needed for antitumor activity in vitro as determined by proliferation assays with four GBM cell lines and also lower compared with previously reported drug concentrations in other tumors (Supplementary Table S5; refs. 16, 19). These findings are in accordance with the publicly available data on sunitinib sensitivity of the Sanger Institute (Hinxton, England). In their experimental design of studying cancer cell line sensitivity to sunitinib, it was found that the geometric mean of the IC₅₀ for sunitinib was 11.6 μmol/L for 398 cancer cell lines, while for the 11 GBM cell lines a higher geometric mean of the IC₅₀ of 18.7 μmol/L was detected (https://www.cancerrxgene.org/compound/Sunitinib/5/overview/ic50), supporting our finding that GBM cells are less sensitive to sunitinib than other tumor cells in general.

The 10-fold higher sunitinib concentrations in GBM tumor tissue compared with plasma in these patients are a result of cellular accumulation of sunitinib due to its physico-chemical characteristics (19, 26). The lower sunitinib concentration in CSF compared with tumor tissue and plasma is most likely due to a significantly lower protein concentration in CSF, which we did not measure and should therefore be interpreted with caution. Protein binding of drugs including sunitinib affects the free drug concentration and therefore the free drug concentration is assumed to be a more accurate measure of the potentially available active drug. However, one should realize that it is the intratumoral intracellular unbound drug that needs to interact with the target for its efficacy. Unfortunately, it is yet not possible to quantify intracellular free drug concentrations without interfering with intracellular drug distribution. Our data on drug distribution in patients with GBM provide additional insight into the currently scarcely available data, while being aware of the fact that the translation from preclinical in vitro and clinical plasma concentrations to intratumor concentrations should be taken with caution (27).

While it is known that the BBB is being disrupted by GBM tumor growth, it presumably still forms a barrier for adequate drug delivery. Brain tumor capillaries are composed of continuous and nonfenestrated capillaries and have high expression of drug efflux transporters, resulting in a hurdle for optimal drug delivery (28). Several TKIs have been shown to be both substrates as well as inhibitors (at higher concentrations) of these drug efflux transporters (29). A clear difference between drug availability of TKIs was shown between flank and orthotopic GBM models explaining the difference in response to treatment for several targeted agents (30–32). The GBM sunitinib concentrations we found here are nearly five times lower than the recently reported tumor concentrations in tumor biopsies of patients with other solid malignancies (9.0–9.5 μmol/L) indicative for hampered tissue penetration through the BBB (Supplementary Table S5; refs. 16, 19). Limited data are available on GBM TKI concentrations in patients, but Plotkin and colleagues reported comparable sunitinib concentrations (converted from ng/g to mmol/L) of 0.45 μmol/L (range 0.14–2.28) in tumor tissue and 0.15 μmol/L (range 0.0825–0.24) in the paired blood samples from these patients with GBM (33). The lack of clinical activity of sunitinib in GBM was shown by Neijns and colleagues in which sunitinib-treated patients with a recurrent GBM had no tumor response nor clinical benefit (2). In addition, compared with its concentration in other malignancies, the relatively low sunitinib concentrations in GBM (we found here) provide clinical evidence of prior preclinical findings that the BBB plays a significant role in the resistance of GBM to TKIs (28, 30–32). Taken together, these preclinical and clinical data indicate that GBM is difficult to treat with sunitinib, not only due to BBB-mediated reduced drug penetration but also due to a lower intrinsic sensitivity of GBM (28). Alternative treatment schedules or dosing to reach higher tumor concentrations have not been considered although several clinically approved TKIs have been evaluated for their potential benefit in GBM (34). Based on the above-mentioned findings one may hypothesize that higher dosing of TKIs will lead to increased exposure and thereby may improve antitumor activity, but at the same time such an approach is difficult to implement in the clinic due to increased toxicity (35). Still, some examples of alternative high dosing TKI strategies with manageable toxicity and promising antitumor activity have been reported (36–38). Based on these findings, a phase II/III clinical trial to evaluate this strategy in patients with recurrent GBM (NCT03025893) is ongoing with sunitinib dosing of 700 mg once every 2 weeks. Alternative approaches such as intrathecal TKI injection may also be considered to optimize local exposure, but this has been studied only in a preclinical model with nano-particulated, water-soluble erlotinib (39).

Besides achieved drug concentrations at the target site, the signaling context or network in tumors is essential for TKI efficacy as well (40, 41). In this study, a comparable number of phosphopeptides as were previously reported by Johnson and colleagues from 7 untreated patients with GBM identified including similar phospopeptide-related pathways as Src and Yes (42). Despite the small number of patients, supervised clustering of the GBM tissue tyrosine-phosphoproteome profiles revealed a clear separation between 4 sunitinib-treated versus 7 untreated patients. This was based on a relatively high number of upregulated phosphopeptides in the sunitinib-treated group (108 vs. 23 downregulated phosphopeptides; cut-off P < 0.05), resulting in an EGFR-centered signaling network (Fig. 3). These included phosphopeptides mapped to proteins which may potentially serve as targets for (combination) therapy, such as EGFR and FGFR, but we stress that the small sample size of the 2 groups calls for cautious interpretation of this comparative analysis.

In this small group of patients no relation between tumor concentration and phosphopeptide regulation was found. Several preclinical studies showed differential effects of TKI treatment on orthotopic versus flank implanted tumors and suggested a relation with achieved intratumor drug concentrations (32, 43). Single-cell phosphoproteomics revealed alteration of protein signaling without genetic changes may cause resistance as early as 2.5 days after start of treatment with an mTOR-inhibitor in a patient-derived in vivo GBM model (44).

In the per sample analysis, MAPK10 was identified as the highest ranking kinase in four tumors using INKA analysis (12). This included sample CON-08 with, compared with the other 10 samples, outlier kinase activity of MAPK10 with a very high INKA score suggestive of amplification (Fig. 4). However, no normal DNA was available for validation by whole-exome sequencing. MAPK10, also known as JNK3, regulates several physiologic neuronal functions including apoptosis (45, 46). Moreover, JNK kinases are known to be expressed and activated in the majority of GBMs and their activation is associated with promotion of GBM tumor growth (47, 48). Several selective small-molecule JNK3 inhibitors have been reported (49), but to the best of our knowledge, no studies currently evaluate their activity in GBM. Notable kinase activity of INSR/IGF1R was found in another individual sample. Dual inhibition of these kinases has shown preclinical activity in GBM (50). Taken together, these data show a complex profile of kinase activity in GBM, supporting the potential of (phospho)proteomic analysis for a better understanding of tumorbiology as well as for the identification of targets for (combination) treatment.

In conclusion, we found in patients with GBM clinical evidence for insufficient intratumoral concentrations of sunitinib to exert its antitumor activity, which may partly explain its intrinsic BBB-mediated resistance and may be overcome by high-dosed, alternatively scheduled administration. In parallel, a complex profile of kinase activity in GBM was found, supporting the potential of (phospho)proteomic analysis for the identification of targets for (combination) treatment.

M. Labots reports personal fees from Bristol-Myers Squibb and MSD outside the submitted work. G.J. Peters reports personal fees from Taiho Pharmaceuticals, Clear Creek Bio, Isofol Medical AB, and Ellipses Pharma and grants and personal fees from Rexahn Pharma outside the submitted work. H.M.W. Verheul reports nonfinancial support from Pfizer outside the submitted work. No disclosures were reported by the other authors.

M.E. van Linde: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Labots: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–review and editing. C.G. Brahm: Formal analysis, investigation, visualization, methodology, writing–review and editing. K.E. Hovinga: Investigation, writing–review and editing. P.C. De Witt Hamer: Conceptualization, supervision, investigation, visualization, methodology, writing–review and editing. R.J. Honeywell: Formal analysis, investigation, methodology, writing–review and editing. R. de Goeij-de Haas: Formal analysis, investigation, methodology, writing–review and editing. A.A. Henneman: Formal analysis, investigation, visualization, methodology, writing–review and editing. J.C. Knol: Formal analysis, investigation, methodology, writing–review and editing. G.J. Peters: Supervision, methodology, writing–review and editing. H. Dekker: Data curation, software, formal analysis, investigation, visualization, methodology, writing–review and editing. S.R. Piersma: Data curation, software, formal analysis, investigation, methodology, writing–review and editing. T.V. Pham: Data curation, software, validation, investigation, methodology, writing–review and editing. W.P. Vandertop: Conceptualization, resources, supervision, writing–review and editing. C.R. Jiménez: Conceptualization, resources, software, supervision, validation, investigation, visualization, methodology, writing–review and editing. H.M.W. Verheul: Conceptualization, resources, formal analysis, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing.