Phase I Study of the Combination of Sorafenib and Temsirolimus in Patients with Metastatic Melanoma

The majority of melanomas have a mutation in the RAS-RAF-MEK-ERK signaling pathway, including activating mutations in the BRAF (∼45%) and NRAS (∼20%) genes (1). Whereas activating mutations in BRAF are thought to predominantly activate downstream signaling via mitogen-activated protein/extracellular signal-regulated kinase (ERK) kinase (MEK) and ERK, RAS GTPases activate additional signaling pathways (2). In particular, the PI3K-AKT-mTOR pathway plays a critical role in oncogenic signaling by RAS, and it is a promising combinatorial target for NRAS-mutant tumors (3, 4). The PI3K-AKT-mTOR pathway is also activated in cancer by loss of function of the lipid phosphatases PTEN, which is detected in 10% to 30% of melanomas (5–7). PTEN loss is mutually exclusive with NRAS mutations, but often occurs in melanomas with a concurrent BRAF mutation. Functionally, loss of PTEN complements activating BRAF mutations to transform melanocytes in a murine model (8). Furthermore, activation of PI3K-AKT-mTOR signaling has been implicated in both primary and secondary resistance to BRAF and MEK inhibitors (9–12). Thus, there is a strong rationale to test the clinical effects of combined inhibition of the RAS-RAF-MEK-ERK and PI3K-AKT-mTOR pathways in melanoma.

Sorafenib (also called BAY 43-9006) is a small molecule inhibitor of several protein kinases, including wild-type and mutant BRAF, CRAF, c-KIT, and VEGFRs (13). Sorafenib is effective at inhibiting the growth and survival of melanoma cell lines and xenografts with BRAF or NRAS mutations in preclinical models (14). However, it achieves a clinical response in only 3% of patients (15, 16). One of these clinical studies evaluated the effects of sorafenib treatment on cyclin D1 and Ki67 expression in patient specimens (16). However, to date, the effects of sorafenib treatment on ERK phosphorylation, which correlates with tumor regression with the selective BRAF inhibitor vemurafenib, has only been reported in one melanoma patient (17, 18).

Combined treatment with sorafenib and rapamycin, an mTOR inhibitor, decreases proliferation and invasion and induces apoptosis in BRAF-mutant human melanoma cell lines (19). Synergistic growth inhibition with this combination is observed in human melanoma cell lines with both mutant and wild-type BRAF (20). Clinical testing of mTOR inhibition in melanoma has previously been reported with temsirolimus (also called CCI-779), an mTOR inhibitor structurally related to rapamycin (21). Temsirolimus was well tolerated, but only 1 of 33 metastatic melanoma patients achieved a clinical response. Although previous studies have examined the degree of inhibition of the phosphorylation (activation) of the mTOR effector S6 in circulating cells and head and neck carcinomas, no assessments have been reported in melanoma patients (22, 23).

We have conducted a phase I clinical trial of concurrent treatment with sorafenib and temsirolimus. The primary endpoint of the study was to evaluate the safety and toxicity of this regimen, and to determine the maximum tolerated dose (MTD) for further testing. The secondary endpoint was to determine the clinical activity of this treatment in metastatic melanoma. Patients with accessible tumors underwent optional biopsies before and during treatment with sorafenib and temsirolimus for quantitative proteomic analysis by reverse phase protein arrays (RPPA; refs. 24, 25). The presence of activating mutations in BRAF and NRAS, and effects on P-ERK and P-S6 protein expression, were compared with the clinical efficacy of the combination.

Patients with histologically confirmed stage IV or recurrent or unresectable stage III melanoma of nonuveal origin were enrolled. Major inclusion criteria included: Eastern Cooperative Oncology Group (ECOG) performance status of 1 or less, measurable disease defined by Response Evaluation Criteria in Solid Tumors (RECIST) version 1.0, adequate organ functions, no active or symptomatic brain metastases and no steroid support for palliation, and uncontrolled blood pressure. There was no restriction in the number of prior therapies.

The protocol for this study was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center, and all patients gave written informed consent before enrollment. The doses of oral sorafenib and intravenous temsirolimus in the first patient cohort were 400 mg twice a day and 15 mg weekly, respectively (dose level 1). Table 1 lists the dose escalation scheme. One cycle was defined as 4 weeks of treatment. The maximum number of treatment cycles planned was 6; patients who continued to have disease regression were allowed to receive more than 6 cycles, as long as they tolerated treatment well. Intrapatient dose escalation was not permitted.

Toxicity was evaluated according to the U.S. National Cancer Institute Common Terminology Criteria for Adverse Events v3.0 (https://webapps.ctep.nci.nih.gov/webobjs/ctc/webhelp/welcome_to_ctcae.htm). Patients were assessed for safety every 2 weeks during the first 2 cycles and then once every cycle. Complete blood counts with differential counts and serum chemistry were obtained once a week during the first 2 cycles and then every other week. Dose-limiting toxicity (DLT) was defined as grade 4 hematologic toxicity, grade 3 or more nonhematologic toxicity, or any grade 2 nonhematologic, or grade 3 hematologic toxicity requiring a dose reduction or treatment interruption for more than 7 days during the first cycle. Grade 3 to 4 nausea, vomiting, or diarrhea were only considered DLTs if they occurred despite optimal medical management. Grade 3 electrolytes, uric acid, phosphorus abnormalities were not considered DLTs if they were correctable within 1 week.

Responses were evaluated after every 2 cycles of the treatment using computed tomography and/or MRI, as well as by direct ruler measurement for clearly visible cutaneous lesions. Clinical responses were evaluated using the international criteria proposed by the RECIST Committee (26).

Blood samples were collected on consenting patients at the following time points of cycle one: Day 1: pretreatment, 1, 4, 6, 8, 10, and 24 hours after start of administration of sorafenib; Day 15: pretreatment, 1, 4, 6, 8, 10, and 24 hours after treatment. Blood samples were collected in 5-mL sodium heparin tubes and obtained in Clinical and Translational Research Center at The University of Texas MD Anderson Cancer Center. Following collection, they were immediately centrifuged at 1,500 rpm for 10 minutes at 4°C for separation. Plasma samples were aliquoted and stored in labeled cryovials at approximately −70°C until analysis. Plasma sorafenib concentrations were determined by liquid chromatography/tandem mass spectrometry. A Waters Acquity UPLC system (Waters) with a 4°C autosampler was run in tandem with a Waters Quattro Premier mass spectrometer using MassLynx/Quanlynx software for data acquisition and integration. Sorafenib was extracted using a liquid–liquid method by adding 4 mL methyl tert-butyl ether to 100 μL of plasma and vortexed for 5 minutes. The resulting mixture was centrifuged at 2,000 × g for 8 minutes at 4°C and frozen at −80°C for 30 minutes. The resulting supernatant was evaporated to dryness at 40°C under gentle nitrogen gas stream and reconstituted with 100 μL of 50:50, methanol:water (v:v). Chromatographic separation was carried out following a 10 μL injection on a Phenomenex Kinetex C18 column (100 × 2.1 mm, 2.6 μm) over 5 minutes at 0.4 mL/min using 50:50, 0.2% formic acid:methanol (v:v) with a linear gradient to 98% methanol over the first 2 minutes. The monitored transition for sorafenib was 465 > 271 m/z with a peak retention time of 2.52 minutes in positive electrospray ionization mode (ESI+). The calibration curve ranged from 1 to 10,000 ng/mL in single donor human plasma. Pharmacokinetic parameter estimates were generated from individual patient concentration–time data using a noncompartmental model (WinNonLin v5.2; Pharsight Corp.).

After signing informed consent, each patient underwent a tumor biopsy (punch or core-needle biopsy), which was embedded in Optimum Cutting Temperature (OCT) solution, unless frozen or OCT-embedded tumor samples obtained within 3 months before signing informed consent were available. An optional tumor biopsy was conducted within 6 hours after the completion of temsirolimus infusion on day 15 of the first treatment cycle. Hematoxylin and eosin stain (H&E) slides were generated from each tumor, which were reviewed by an experienced dermatopathologist. Tumors with regions of at least 70% viable tumor cells of sufficient size for DNA and protein extraction underwent H&E-guided macrodissection (7). Shears of the tumor-enriched isolates were prepared using a DNAse-free, RNAse-free microtome at −20°C, then were stored at −80°C. DNA was extracted from the samples and underwent mass spectroscopy–based genotyping for mutations in BRAF (exon 15) and NRAS (exons 1and 2), as previously described (7, 27). Proteins were extracted from the samples as previously described (7). Protein lysates were normalized to the same concentration, denatured, and then submitted for RPPA analysis through the MD Anderson Cancer Center RPPA Core Facility (http://www.mdanderson.org/education-and-research/resources-for-professionals/scientific-resources/core-facilities-and-services/functional-proteomics-rppa-core/index.html). Log₂ expression data for each protein was corrected for measured differences in total protein loading as previously described (7). A detailed description of the RPPA procedure is provided in the Supplementary Materials.

The MTD was defined as the highest dose at which the incidence of DLT was less than 33%. A conventional 3+3 dose escalation design with predefined dose levels of sorafenib and temsirolimus was implemented. At the MTD, a total of 10 patients were enrolled to ensure that the safety profile was acceptable. Overall clinical response included both complete and partial responses. Progression-free survival (PFS) and overall survival (OS) were measured from day 1 of treatment. One patient discontinued the study due to toxicity before disease progression and used different therapies; this patient was censored for PFS at the time of study discontinuation. Patients that did not experience death were censored at the time of their last clinic follow-up. The duration of clinical response was measured from the time a clinical response was first achieved until the time of disease progression or death, whichever occurred first.

The association between tumor mutation status and clinical response was determined using Fisher exact test. Kaplan–Meier survival curves were used to estimate PFS and OS; log-rank testing was used to compare curves. Pharmacodynamic effects of treatment with sorafenib and temsirolimus on P-ERK and P-S6 were determined by calculating the ratio of phospho:total protein expression in each sample. Pretreatment and day 15 ratios were compared using paired Student t test.

Twenty-five patients were enrolled and treated in the study (Table 2). The median age of the patients was 51 years. Forty percent of the patients had stage III melanoma, 12% had stage IV M1a, 28% stage IV M1b, and 20% stage IV M1c, respectively. Fifty-six percent of the patients had received at least 2 prior systemic therapies for metastatic melanoma.

One of the first 3 patients treated at dose level 1 had a DLT (grade 3 hand-foot syndrome and thrombocytopenia requiring dose reduction; Table 3). However, no other DLTs were observed among 3 patients treated at dose level 0, nor among an additional 3 patients treated at dose level 1. At dose level 2, both of 2 patients had a DLT (grade 3 serum transaminase elevation and hand-foot syndrome). To better define a MTD of the combination, 2 additional dose levels (level 1a and 1b) were evaluated. All 4 patients at dose level 1a tolerated the treatment without DLTs. The next 6 patients were treated at dose level 1b with only 1 patient experiencing reversible grade 3 hypertriglyceridemia. Four additional patients were treated at dose level 1b, and 2 of them experienced DLTs of grade 3 or more thrombocytopenia requiring dose reduction. Dose level 1b (sorafenib 400 mg in am and 200 mg in pm daily and temsirolimus 25 mg weekly) was selected as the MTD.

The most common grade 3 or 4 nonhematologic adverse events were fatigue and hypophosphotemia (Table 4). The most common grade 3/4 hematologic toxicities were thrombocytopenia (n = 3) and lymphopenia (n = 3; Table 4). Seventeen (68%) patients required dose reduction during the course of the treatment. All adverse events resolved upon treatment discontinuation.

The plasma concentration–time profiles of sorafenib were characterized by a relatively slow absorption and elimination phase with primary and secondary peaks in a number of patients, as seen in a previous single-agent phase 1 study (18). Sorafenib mean peak plasma concentrations (C_max) and mean daily exposure, measured as the average concentration over 24 hours (C_avg), was 2- to 4-fold greater at steady state on day 15 when compared with day 1 with low-to-moderate interpatient variability across all dosing levels (Table 5). Sorafenib mean C_avg at day 15 with 400 mg twice daily dosing was 8.45 mg/L, which was more than a dose proportional increase when compared with 400 mg every day (3.16 mg/L); mean C_max increased proportionately (11.83 vs. 5.38 mg/L). Mean C_avg with sorafenib dosing of 400 mg am/200 mg pm on day 15 was 4.55 mg/L, which was similar to the result observed in the one patient treated with 200 mg twice daily with available day 15 data (C_avg 5.12 mg/L). The sorafenib mean C_max was similar for all dose levels on day 1 but became more dose proportional by day 15 (except for 200 mg twice daily). This trend and with the mean C_max levels at 400 mg twice daily were similar to those observed in a review of 4 phase I trials with single-agent sorafenib in patients with advanced cancer (28).

None of the patients had a clinical response. Ten patients had disease stabilization for at least 8 weeks; only 3 patients had disease stabilization for more than 16 weeks. The median number of cycles received per patient was 2 (range < 1–8). Overall the median PFS was 2.1 months (Fig. 1), and the median OS was 1.3 years (Supplementary Fig. S1). Among the 10 patients treated at the MTD, 2 achieved stable disease, the median PFS was 2.0 months, and the median OS was 1.2 years.

BRAF and NRAS tumor mutation status was determined for the patients. Ten (40%) patients had activating BRAF mutations, 5 (20%) had NRAS mutations, 8 (32%) had wild-type BRAF and NRAS genes; mutation status could not be determined in 2 patients (8%). There was no significant difference in median PFS (P = 0.34, Fig. 1B) or OS (P = 0.98, Supplementary Fig. S1) by mutation status. Tumor mutation status was also not associated with the chance of achieving disease stabilization as a best response (P = 0.75, Supplementary Table S1). A review of patient records identified 9 patients with clinical testing for c-KIT gene mutations. One patient, who had wild-type BRAF and NRAS genes, was found to have a c-KIT L576P mutation. This patient had stable disease as the best response and remained on treatment for 8 cycles before progressing, which was the longest PFS of any patient in the trial. None of the other patients with c-KIT testing results had a mutation in the gene.

Ten patients with pretreatment tumor biopsies also underwent an optional tumor biopsy on day 15 of cycle 1. Four patients underwent paired core needle biopsies of lymph node metastases, and 6 underwent paired punch biopsies of cutaneous metastases. Review of the core needle biopsies showed that none of the samples contained isolates of more than 70% viable tumor of sufficient size for the planned proteomic analyses, whereas 4 of the 6 pairs of punch biopsies were sufficient for RPPA. Two of the tumors were BRAF-mutant (1 V600E, 1 V600K), 1 was NRAS-mutant (Q61K), and 1 was wild type for both genes. The NRAS mutant had stable disease as best response and received 4 cycles of treatment; the other 3 patients had progressive disease and received only 2 cycles. Comparison of the levels of the ratio of phospho:total S6 protein showed that all 4 paired samples had at least a 25% decrease, and 2 had at least 75% decrease, in P-S6 following treatment with sorafenib and temsirolimus, consistent with inhibition of mTOR by temsirolimus (Fig. 1C). In contrast, only 1 of the 4 samples showed a decrease of phospho:total ERK levels of greater than 25% (wild-type, progressive disease; Fig. 1D).

Clinical experiences in multiple cancers have shown the promise of targeted therapies. However, these experiences have also identified the frequent, and often rapid, emergence of resistance as an important challenge (29). On the basis of pattern of molecular alterations observed in melanoma and the results of preclinical studies, we carried out a phase I clinical trial to determine the efficacy and toxicity of combined treatment with sorafenib, an inhibitor of BRAF and other kinases, and temsirolimus, an mTOR inhibitor, in patients with advanced melanoma. Our study identified significant toxicity with the combination at higher doses and failed to achieve clinical responses in any of the patients. Biomarker analyses failed to identify any relationship between activating mutations in the RAS-RAF-MEK-ERK signaling pathway with PFS. Furthermore, pharmacodynamic analysis showed that the combination achieved minimal inhibition of ERK activation.

Sorafenib was the first BRAF inhibitor to undergo clinical testing in melanoma following the discovery of frequent activating BRAF mutations in this disease. Although it had shown promising results in preclinical models, only 1 of 37 metastatic melanoma patients achieved a clinical response with sorafenib monotherapy (15). Recently, the mutant-selective, second-generation BRAF inhibitors vemurafenib (also known as PLX4032 or RG7204) and GSK2118436 have shown dramatic efficacy in metastatic melanoma patients with BRAF mutations (29, 30). In the phase III trial comparing vemurafenib to dacarbazine, treatment with vemurafenib resulted in significant improvement in the clinical response rate, PFS, and OS (31). Analysis of patients treated with vemurafenib in a phase I study showed that clinical responses correlated with achievement of more than 80% inhibition of P-ERK expression in the tumor cells (17). Preclinical studies showed that sorafenib inhibits P-ERK in human melanoma cell lines with activating BRAF mutations in vitro (14). In contrast to both the vemurafenib trial and the preclinical studies, in this trial of combined treatment with sorafenib and temsirolimus quantitative analysis of P-ERK levels in 4 patients with evaluable matching pretreatment and day 15 biopsies showed less than 10% inhibition in 3 of the patients, and 46% inhibition in one. Pharmacokinetic analysis found that the maximum and average levels of sorafenib on day 15 detected in this trial (Table 5) are similar to those previously reported in sorafenib single-agent phase I trials (32). Thus, it is unlikely that the failure of sorafenib to inhibit P-ERK is due to a pharmacokinetic interaction with temsirolimus. Although the in vivo pharmacodynamic failure of sorafenib to inhibit the RAS-RAF-MEK-ERK pathway has been hypothesized to contribute to its lack of efficacy, the results of this trial provide direct evidence to support this conclusion. These findings are consistent with an analysis of matched specimens from melanoma patients treated with sorafenib that did not identify significant changes in Ki67 or cyclin D1 expression, although these are not direct pharmacodynamic readouts of RAF kinases (16). Another study reported a positive correlation between pretreatment expression levels of VEGFR2 and clinical responsiveness to treatment with sorafenib, paclitaxel, and carboplatin (33). VEGFR2 is a known target of sorafenib, and likely it is responsible for its clinical activity in renal cell carcinoma. In all 3 studies, the presence of activating BRAF mutations failed to correlate with clinical benefit from sorafenib. Taken together, the data supports that sorafenib should not be used to inhibit the RAS-RAF-MEK-ERK pathway in melanoma patients, particularly those with activating BRAF mutations.

In contrast to sorafenib, the measurement of P-S6 levels in the biopsies supports that temsirolimus was able to inhibit mTOR signaling at doses that were tolerated in this trial. This result is consistent with clinical studies with temsirolimus in other cancers (22, 23). Although mTOR inhibitors have been well tolerated as single agents, they have failed to show significant clinical activity in several solid tumors, including melanoma (21). Preclinical studies showed that inhibition of the mTORC1 complex, which occurs with rapalogs such as temsirolimus, disrupts a negative feedback loop, resulting in hyperactivation of AKT (34). Consistent with this, biopsies from patients treated with everolimus, another mTOR inhibitor, showed a decrease in P-S6, but an increase in P-AKT Ser473 (35). In our study, there was no increase in P-AKT Ser473 levels with combined treatment with temsirolimus and sorafenib (Supplementary Fig. S2). It is possible that concurrent treatment with sorafenib affected this feedback loop, as there seem to be multiple interactions between the RAS-RAF-MEK-ERK and PI3K-AKT-mTOR signaling pathways (10, 36, 37). However, a similar lack of increase in P-AKT Ser473 following single-agent temsirolimus was also recently observed in matching biopsies in head and neck cancer patients (23).

Although there was no significant association between BRAF or NRAS mutations and disease stabilization or PFS, the patient in the trial with the longest treatment duration (8 months) was found to have a c-KIT L576P mutation. Two prior reports have described clinical responses and symptomatic benefit from sorafenib in metastatic melanoma patients with c-KIT V560D (38) and c-KIT D820Y (39) mutations, respectively. Interestingly, a biopsy done after progression on sorafenib in this patient identified the continued presence of the c-KIT L576P mutation and failed to identify any secondary mutations (40). As secondary resistance to KIT inhibitors in melanoma patients with c-KIT mutations is frequent, and as yet cannot be attributed to the presence of additional mutations in the c-KIT gene, it is possible that combinatorial approaches involving inhibitors targeting other pathways may be clinically beneficial in these patients.

In conclusion, we report here that treatment with sorafenib and temsirolimus is toxic and provides minimal clinical benefit in patients with advanced melanoma. This regimen was developed to test the antitumor effects of combined inhibition of the RAS-RAF-MEK-ERK and the PI3K-AKT-mTOR signaling pathways. On the basis of our pharmacodynamic analysis, we do not believe that this regimen achieved enough inhibition of the RAS-RAF-MEK-ERK signaling pathway to test the clinical benefit of such a therapeutic approach. A number of additional potent inhibitors against each of these pathways have become available for clinical testing since the design of this trial, and, likely, will be tested in combinatorial approaches in the near future. Our results, and the studies that led to the development of clinically effective BRAF inhibitors (17), support the critical need for pharmacodynamic evaluations to appropriately interpret the results of such trials.

M.A. Davies received commercial research grants from GlaxoSmithKline, AstraZeneca, Merck, and Roche. He is a consultant and on the advisory board of GlaxoSmithKline. K.B. Kim is a consultant and on the advisory board of Bayer and Onyx.

This work was supported in part by NCI grants UO1 CA062461 and N01 CM17003, the NIH through the MD Anderson Cancer Center support grant CA016672, the P50 SPORE Career Development Grant (K.B. Kim), the ASCO Young Investigator Award (M.A. Davies), and the MD Anderson Cancer Center Physician-Scientist Program (M.A. Davies, A.J. Lazar). The MDACC RPPA Core Facility is supported by M. D. Anderson Cancer Center Core Grant NIH-NCI-CA-16672.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.