Phase II Clinical Trial of Everolimus in a Pan-Cancer Cohort of Patients with mTOR Pathway Alterations

mTOR is a serine–threonine kinase that is the key component of two multi-subunit complexes with non-overlapping downstream targets: mTOR Complex 1 and 2. mTORC1 plays a vital role in the regulation of cell growth and division by controlling the balance between anabolic and catabolic cellular processes, and facilitates cell-cycle progression from G₁ to S phase (1, 2). Given its key role in anabolic processes needed for preparation for cell division, it is not surprising that the majority of cancer cells have activation of mTORC1 through one mechanism or another (3, 4). Such activation may occur downstream of the PI3K/AKT or Ras/MEK/ERK pathways, or alternatively through inactivating mutations in its negative regulators, including TSC1/TSC2 and PTEN (5).

Everolimus, a derivative of rapamycin, binds to FKBP12 to form a complex that is an allosteric inhibitor of mTORC1 (6). Everolimus has been FDA approved for the treatment of renal cell carcinoma (RCC; ref. 7), pancreatic neuroendocrine tumors (8), and hormone-positive HER-2–negative breast cancer in postmenopausal women (9). Everolimus and other rapalogs have also shown activity in multiple neoplasms occurring in the autosomal dominant disorder tuberous sclerosis complex, due to germline-inactivating mutations in TSC1 or TSC2, including renal angiomyolipoma, lymphangioleiomyomatosis (LAM), and subependymal giant cell astrocytoma (10–12). Two hit loss of either TSC1 or TSC2 occurs in these TSC-associated tumors, leading to constitutive activation of mTORC1, because the TSC protein complex functions as a critical GTPase-activating protein for RHEB, which regulates mTORC1 activity.

A previous report on the potential benefit of everolimus in bladder carcinoma indicated that there was a single exceptional responder among 37 treated patients, and this responder had tumor mutations in both TSC1 and NF2 (3). Several other case reports and small series have also reported on exceptional responses to rapalog therapy in occasional patients with TSC1 or TSC2 mutations (13–15). These findings prompted us to carry out a pan-cancer “basket” study of everolimus for treatment of cancers of any histology that had an inactivating mutation in TSC1 or TSC2 or activating mutation in MTOR. In this cohort of 30 subjects, we sought to identify other factors that might play a role in determining response to everolimus, including pathology, clinical features, and co-occurring mutations in other genes.

Eligible patients had advanced malignancies of any histology, with inactivating mutations in TSC1 or TSC2 or activating mutations in MTOR, identified on recent biopsies (within 2 years) of primary or metastatic lesions. Disease had to be measurable by clinical or radiologic examination, with at least one lesion that had diameter ≥10 mm on CT scan, MRI, or calipers by clinical exam. Patients had to be ≥18 years old, able to give informed consent, Eastern Cooperative Oncology Group (ECOG) performance status 0 or 1, and have normal organ and marrow function. Patients may have received any prior therapies with the exception of PI3K or mTOR inhibitors in non-adjuvant settings. Patients were excluded if there were recent other treatments, including chemotherapy within the last 2 weeks, and radiotherapy or investigational drugs within the last 3 weeks; uncontrolled brain or leptomeningeal metastasis; or uncontrolled intercurrent illness.

This was a prospective, single-arm study conducted at two sites [Dana-Farber Cancer Institute (DFCI) and Memorial Sloan Kettering Cancer Center (MSKCC)], was approved by human research committees at each institution, and was conducted in accordance with recognized ethical guidelines (the Belmont Report, U.S. Common Rule). Written informed consent was obtained from all subjects. It was designed to assess the efficacy and toxicity of everolimus in patients with advanced solid tumors of any histology, with confirmed inactivating mutations in TSC1 or TSC2 or activating mutations in MTOR identified in any Clinical Laboratory Improvement Amendments (CLIA)-certified laboratory. Mutations in these three genes were assessed for pathogenicity in each case by study chair (D.J. Kwiatkowski).

This was an investigator-initiated trial for which funding support, in part, was provided by Novartis Inc. The funder had no role in the selection of patients, any aspect of patient management, or the analysis and preparation of this article. This trial was registered at clinicaltrials.gov as NCT02201212. Power analysis was performed to calculate the number of patients needed to detect the targeted response rate of 30%. We had a planned sample size of 30 patients, which was met. There was no role for randomization in this single-arm trial. For the same reason, blinding was not relevant to this study.

Subjects took 10 mg oral everolimus daily, and a treatment cycle was defined as 28 days or 4 weeks of treatment. Dose reductions to 5 mg daily and 5 mg every other day were permitted when there was toxicity. Treatment continued indefinitely or until there was: disease progression, unacceptable toxicity, lack of compliance with the regimen, serious intercurrent illness, or if the patient decided to withdraw from the study for any reason. Treatment beyond progression was permitted when the treating physician felt there was clinical benefit.

At study entry, assessment included: Screening for inclusion/exclusion criteria, medical history, weight and vital signs, ECOG performance status, physical examination, pregnancy test (for females <50 years of age), hematology and blood chemistry, urine analysis, and CT/MRI of chest, abdomen and other disease sites. Subjects were reevaluated for response every 8 weeks by physical examination and repeat scans of areas of known disease and other sites as clinically indicated.

Objective response was evaluated using unidimensional measurements of lesions following RECIST 1.1 (16).

Safety assessments were conducted every 4 weeks. Grading of adverse events was based on the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE v4.0; ref. 17).

All statistical analyses were performed using Statistical Analysis Software (SAS). The primary objective of this trial was objective response rate (ORR) according to RECIST 1.1 criteria. Secondary objectives included clinical benefit rate, duration of clinical benefit, progression-free survival (PFS), and overall survival (OS). Clinical benefit was defined as complete response, partial response (PR), or stable disease (SD) with some degree of tumor shrinkage by RECIST measurements lasting more than 22 weeks, consistent with previous reports (7, 18, 19). Duration of clinical benefit was defined as the time between first documented clinical benefit (per RECIST 1.1) until the time of disease progression. PFS was defined as the time from study registration to documented disease progression (clinically, or as per RECIST 1.1) or death from any cause. Patients not experiencing a PFS event were censored at the date of their last disease assessment. OS was defined as time from registration to death from any cause, censoring patients alive at the last date of follow-up.

Adverse events were tabulated and reported according to category and grade. We estimated event-time distributions using the Kaplan–Meier method, using Greenwood's formula to estimate the variance and provide 95% confidence intervals for the medians.

Tumor genomic profiling was performed for all 30 subjects, using either archival tissue or fresh-frozen biopsies. Initial tumor profiling for study entry was performed by either: OncoPanel [ref. 20; Department of Pathology at Brigham and Women's Hospital (Boston, MA)], MSK-IMPACT [refs. 21, 22; MSKCC (New York, NY)] or Foundation One (Foundation Medicine; https://www.foundationmedicine.com), all of which are gene panel exome sequencing platforms of size 300–500 genes. Subsequently, whole-exome sequencing (WES) was performed for 24 subjects at the Broad Institute of MIT and Harvard (Cambridge, MA) and analyzed using a standard analytic pipeline deployed in the Firecloud/Terra environment (https://terra.bio). WES data were analyzed in both paired “tumor-normal'' or unpaired “tumor-only” mode, to identify single nucleotide variants (SNV), insertions/deletions (indels), and copy-number alterations. Clonality assessment of TSC1/TSC2 and MTOR variants was performed on the basis of cancer cell fraction values generated by ABSOLUTE for WES variant calls and/or variant allele frequencies in relation to the tumor purity. See Supplementary File and Supplementary Table S1 for details on patients' samples characteristics and processing, WES methodology and quality metrics, computational tools for somatic mutation identification, and criteria for genomic data analysis and review.

Thirty patients were enrolled between March 2015 and October 2018, 11 at DFCI and 19 at MSKCC (Supplementary Table S2.1). Baseline demographic and clinical characteristics of the patients are presented in Table 1 (and Supplementary Table S2.1). Patients were 50% female with median age at enrollment of 61.5 (range, 43–80) years, and 83% had ECOG score of 1. Patients had 15 different primary tumor histologies, the most common being bladder cancer (6 of 30; 20%; Fig. 1A). The median number of prior systemic treatment regimens received was 2.5 (range, 0–8; Supplementary Table S2.2).

There were 100 reported adverse events overall (Supplementary Tables S2.4 and S3), which were similar to previous experience with everolimus. Six were grade 3 or higher, including one case of fatal pneumonitis.

Twenty-two of 30 patients were assessed for objective response by RECIST v1.1 eight weeks after enrollment (Fig. 1B; Supplementary Table S2.3). The 8 remaining patients were off treatment at the time of the first radiographic assessment: 4 patients had died, 3 had overt clinical progression such that they were removed from the study prior to 8 weeks, and 1 was non-compliant with the treatment regimen. The ORR to everolimus was 7% (95% CI, 1%–22%): 2 of 30 patients in the intent-to-treat analysis (Fig. 1C). Eleven additional patients had some degree of tumor shrinkage at 8 weeks or later follow-up, and 12 of 30 (40% of the intent-to-treat population) had SD at the time of the initial follow-up assessment. Two subjects who did not meet PR criteria had decreases in absolute measured tumor burden sustained for at least 22 weeks. Thus, the protocol-defined clinical benefit rate was 4 of 30 patients (13%; 95% CI, 4%–31%). The duration of clinical benefit ranged between 1.8 and 15.2 months (Supplementary Fig. S1A), with 1 patient maintaining PR through 21 months and clinical benefit until 27 months after start of treatment (Supplementary Table S2.1). The median PFS was 2.3 months (95% CI, 1.8–3.7 months; Supplementary Fig. S1B), whereas the median OS was 7.3 months (95% CI, 4.5–12.7 months; Supplementary Fig. S1C).

Recognizing that we observed a disappointing response rate, we carefully reviewed the mutations that had been identified as entry criteria. Fifteen subjects had inactivating mutations in TSC2, 13 in TSC1, 1 subject had concurrent inactivating TSC1 and TSC2 mutations, and 1 subject had an activating mutation in MTOR (Table 2; Supplementary Table S4; Fig. 2). Indels and nonsense mutations together constituted the majority of the mutations in both TSC1 and TSC2 [12 of 14, 86%; and 11 of 17 (1 patient with two TSC2 mutations), 65%; respectively; Fig. 2]. There were no recurrent mutations in either gene. One activating mutation in MTOR, c.6644C>T (p.Ser2215Phe), a known hotspot gain-of-function mutation, was seen (22). Twenty-three of 24 (96%) mutations for which clonality could be assessed were clonal (Supplementary Table S4). TSC1/TSC2 biallelic inactivation was evident in 18 of 23 (78%) subjects in this cohort, and was due to copy-number loss of heterozygosity (LOH) in 16 of 18 (89%). P21 had two distinct TSC2 somatic point mutations in the tumor, whereas P24 had loss of both copies of TSC2 (Table 2; Supplementary Table S4).

Although the numbers of subjects with response was quite small, there did not appear to be a correlation between gene or type of mutation with response (Figs. 1B and C, 2A). Two subjects with TSC2 mutation and two with TSC1 mutation had clinical benefit from everolimus. Three of 4 patients with clinical benefit had biallelic inactivation of either TSC1 or TSC2; for the remaining 1 patient the assessment of LOH status/biallelic inactivation was not possible. This is similar to the overall cohort.

We examined whether other genomic events might be associated with response to everolimus beyond TSC1/TSC2 and MTOR (Fig. 3). WES was performed on all subjects with available tumor material, n = 24, including 19 with available normal tissue for tumor-normal comparison, and 5 that were unpaired (Supplementary Table S1). When WES was not available, we included mutation data from the panel sequencing performed for study entry (Supplementary Tables S5 and S6). WES confirmed the TSC1/TSC2/MTOR mutations identified in panel sequencing in all 24 cases.

Tumor mutation burden (TMB) for both nonsynonymous and all mutations were highly variable in the 19 subjects with paired tumor-normal sequencing, and was not associated with clinical benefit in an obvious manner (Supplementary Fig. S2; Supplementary Table S1). TP53 was the most commonly mutated gene in this multi-histology cohort, with 16 nonsynonymous mutations in 15 of 30 subjects (50%), and did not appear to be associated with clinical response (Fig. 3). The TERT promoter, with two oncogenic hotspots: g.1295228C>T and c.1295250C>T (https://www.cbioportal.org; ref. 21) was mutated in 11 of 25 patients (44%; for whom the TERT promoter genomic region was covered; Fig. 3). Interestingly, 75% (9 of 12) of tumors with inactivating TSC1 mutations, had concurrent TERT promoter mutations, whereas this was seen in only 2 of 11 (18%) subjects with TSC2 mutations (P = 0.01, Fisher exact test). There was, however, no association between TERT mutations and response.

The 2 subjects who achieved PR had uterine (P3) and upper tract urothelial carcinoma (UTUC; P18), respectively; TSC2 and TSC1 mutations, respectively; and both had concurrent KMT2C and SMARCA4 nonsynonymous mutations/two-copy deletions (Fig. 3). Tumor P3 had two-copy deletion of SMARCA4, and a missense KMT2C mutation [variant of unknown significance (VUS), p.Ile4780Thr] without loss of the second KMT2C allele. Tumor P18 had a truncating KMT2C mutation (p.Gln2161*) with uncertain status for the second allele, and a missense VUS in SMARCA4 (p.Phe1142Leu). Tumor P18, UTUC, had an exceptionally high nonsynonymous point mutation rate (12.28 mutations/Mb), higher than any other sample in this cohort, and mutation signature analysis showed it was likely due to APOBEC activity (C>G or T at TCW), commonly seen in urothelial cancers (23). Multiple nonsynonymous mutations were seen in DNA damage response (including TP53 and BRCA2) and chromatin remodeling genes (ARID1A, KMT2C, and KMT2D), reported to be frequently mutated in urothelial cancers enriched with APOBEC signature mutations (23–25). This subject displayed the greatest response, near 50% reduction by RECIST criteria, but also developed new sites of disease at 5 months on treatment.

Two subjects with clinical benefit [P18 (UTUC, PR) and P22 (anal cancer, SD>22 weeks)] had concurrent nonsynonymous mutations in both KMT2C and KMT2D genes, in addition to TSC1 mutations. P18 had a biallelic inactivation of TSC1, along with (i) two nonsynonymous KMT2D mutations [missense (VUS, p.Glu3081Lys) and truncating (p.Ser4727*)], and (ii) a truncating KMT2C mutation (p.Gln2161*). P22 had a nonsense TSC1 mutation with uncertain status for the second allele and concurrent (i) two truncating KMT2D mutations (p.Pro3659Valfs*88 and p.Cys346Serfs*17) and (ii) three nonsynonymous KMT2C mutations [including 2 VUS missense (p.Ser2053Cys and p.Leu2036Val) and a truncating mutation (p.Ser2059*)].

Among the 4 subjects with clinical benefit, 2 subjects had uterine leiomyosarcoma (P3 and P21), and both had mutations in each of TSC2 and NF1 (Fig. 3). P3 had an NF1 truncating mutation (p.Gln2138*) with one copy deletion of NF1 exon 1–20, whereas P21 had a clonal NF1 missense mutation [p.Arg1830Ser, pathogenic according to ClinVar (26) and VarSome (27)] along with deletion of one allele of NF1. Both the NF1 SNVs and the deletions were seen in both the WES and the targeted MPS OncoPanel analyses. These findings suggest the possibility that biallelic concurrent NF1-TSC2 inactivation may sensitize uterine leiomyosarcoma to everolimus response. However, biallelic concurrent NF1-TSC2 inactivation was also seen in P1 (ovarian cancer) who did not achieve clinical benefit (best response: SD; Fig. 3).

None of the patients in our cohort had mutation in NF2, seen previously in a patient with extraordinary response to everolimus (tumor shrinkage of >80% over 23 months of everolimus treatment), with metastatic bladder cancer and concurrent TSC1 mutation (3).

The TSC protein complex, consisting of TSC1, TSC2, and TBC1D7 is well known to be a critical regulator of the state of activation of the mTORC1 kinase complex, due to its function as a GTPase-activating protein for the RHEB GTPase, which plays a critical role in regulation of mTORC1 (28, 29). Complete loss of either TSC1 or TSC2 in a variety of cellular and in vivo models leads to constitutive activation of mTORC1. Rapalogs allosterically inhibit the kinase activity of mTORC1, through binding to FKBP12, and then binding of the rapalog–FKBP12 complex to the FRB domain of mTORC1 (6). However, rapalogs do not inhibit the kinase activity of mTORC1 to an equal extent among all its substrates, with S6K being among the most sensitive substrates, and 4EB-P1 among the most resistant (30).

The overall response rate in this trial was clearly disappointing. Our results indicated that inactivating mutations in TSC1/TSC2 are not a predictive biomarker for response to everolimus therapy for advanced malignancies in general. It is appropriate to consider these results in the context of prior reports. In TSC-related tumors, occurring both in individuals with the TSC syndrome and those without TSC, such as renal angiomyolipoma, subependymal giant cell astrocytoma, cardiac rhabdomyoma, and pulmonary LAM, rapalogs have shown benefit in well over 50%, and are the mainstay of treatment (8–12). However, although responses are durable in these patients, including re-response when therapy is discontinued and re-started, the degree of response is modest, and few patients achieve a PR by formal RECIST criteria. In malignant perivascular epithelioid cell tumors (PEComas), a set of tumors with histologic similarity to angiomyolipoma that occur sporadically, and for which TSC1/TSC2 mutation is seen in about 50%, response to rapalog therapy is also commonly seen (31, 32). In other solid tumors, rare exceptional responders to rapalog therapy have been reported, who have had either TSC1/TSC2 or MTOR mutations (13–15). Rapalogs are FDA approved for treatment of RCC, but the evidence regarding the association between mTOR pathway mutations and response in RCC is weak, with outlier analyses suggesting correlation in extreme responders, and studies in unselected cohorts showing no relationship (18, 33). The mechanism of benefit from rapalogs in RCC is uncertain, but may relate to inhibition of HIF-1a (34, 35).

Everolimus is also FDA approved for the treatment of postmenopausal women with advanced hormone receptor–positive, HER2-negative breast cancer in combination with exemestane (9). However, it is notable that the CR+PR rate to the combination was only 9.5%, versus 0.4% for exemestane alone, whereas there was a significant improvement in PFS from 2.8 months to 6.9 months. The mechanism of benefit in this clinical circumstance is not clear but may be due to interactions between ER and mTORC1 signaling, leading to a synergistic effect of the combination of inhibitors. TSC1/TSC2/MTOR mutations are all very rare in breast cancer.

In this trial, PRs were seen in 2 of 30 (7%) patients. Two additional patients [total 4 of 30 (13%) patients] had treatment effects that met the protocol-specified definition of clinical benefit. Notably, 13 patients had some degree of tumor shrinkage by RECIST on their initial follow-up scan, but for 9 of those patients, this finding was transient. Although these changes in tumor burden measurements by CT imaging are within typical linear measurement interassessment variance, some element of the observed reductions could be due to effects of everolimus on tumor cell size, a consistent effect seen with mTOR inhibition in vitro (36, 37). The transient nature of this tumor shrinkage may reflect rapid adoption of bypass or resistance mechanisms in this relatively large set of minimally responsive tumors.

With the extremely low response rate, any observations on association with response are considered exploratory. However, there was no association between gene with mutation, or concurrent mutations, and response to everolimus. Two patients with TSC1 and 2 with TSC2 mutations had protocol-defined clinical benefit from everolimus. Of these 4 patients, 2 (P3 – PR; P21 – SD for >22 weeks) had concurrent biallelic TSC2 and NF1 inactivation, suggesting that TSC2-NF1 co-mutation may sensitize tumors to everolimus. Both of these 2 responders with TSC2-NF1 co-mutation had high-grade uterine leiomyosarcoma. The tumor from the patient with PR, P3, had pathologic features that were PEComa-like, including staining for MITF and Cathepsin K, whereas the other did not have such features. It is possible that the PEComa characteristics of the P3 tumor may explain the good response seen in that patient. Three other patients with leiomyosarcoma were included in this trial, 2 of which dropped out early, and 1 had progressive disease as best response. Hence, additional factors, including tumor lineage, heterogeneity, co-alteration profile, and epigenetic state may impact degree of response to everolimus therapy.

One limitation of our study was that a single metastatic site was sequenced in all patients (1 patient had the primary site sequenced as well). Tumor heterogeneity may have led to lack of consistency of TSC1/TSC2/MTOR mutation findings in different tumor sites, leading to lack of response to everolimus.

In conclusion, this pan-cancer prospective trial evaluating everolimus therapy for TSC1/TSC2/MTOR mutation-bearing cancers was disappointing with a low ORR (7%). For the less stringent, protocol-defined detection of potential therapeutic effect, 13% had “clinical benefit.” With this low response rate, all clinical and molecular features found to be associated with response must be considered exploratory and tentative. However, there was no association between gene with mutation, or presence of bi-allelic inactivating mutation with response. There is a suggestion of enhanced response in those subjects with concurrent NF1-TSC2 mutations, and in those with pathologic features of PEComa; this observation is considered preliminary and requires further investigation in independent cohorts. The development of more effective mTORC1 inhibitors will hopefully result in higher rates of durable responses in patients with alterations in this pathway.

L.M. Sholl reports grants from Genentech, as well as personal fees from AstraZeneca and EMD Serono outside the submitted work. D.M. Hyman reports personal fees from Loxo Oncology, and a wholly owned subsidiary of Eli Lilly during the conduct of the study. G.I. Shapiro reports other from Novartis during the conduct of the study. G.I. Shapiro also reports grants from Eli Lilly and Merck & Co., grants and personal fees from Merck KGaA/EMD-Serono and Sierra Oncology, as well as personal fees from Pfizer, G₁ Therapeutics, Roche, Bicycle Therapeutics, Fusion Pharmaceuticals, Cybrexa Therapeutics, Astex, Almac, Ipsen, Bayer, Angiex, Daiichi Sankyo, Seattle Genetics, Boehringer Ingelheim, ImmunoMet, Asana, Artios, Atrin, Concarlo Holdings, Syros, Zentalis, and CytomX Therapeutics outside the submitted work. J.J. Harding reports grants and personal fees from Bristol Myers Squibb, as well as personal fees from Merck, Adaptimmune, Exelexis, Eisai, Eli Lilly, QED, Zymeworks, Imvax, and CytomX outside the submitted work. M.H. Voss reports personal fees from Pfizer, Corvus, and Exelixis; grants from Pfizer; and personal fees from Eisai, Merck, Calithera, and Aveo outside the submitted work. G. Iyer reports grants from Novartis during the conduct of the study. G. Iyer also reports grants and personal fees from Mirati Therapeutics, personal fees from Basilea Pharmaceutica, grants and personal fees from Janssen, and grants from Bayer and DeBioPharm outside the submitted work. D.J. Kwiatkowski reports personal fees from AADi and grants from Revolution Medicines outside the submitted work. No disclosures were reported by the other authors.

The funder had no role in the selection of patients, any aspect of patient management, or the analysis and preparation of this article.

E. Adib: Data curation, formal analysis, visualization, methodology, writing–original draft, writing–review and editing. K. Klonowska: Data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. K. Giannikou: Data curation, writing–review and editing. K.T. Do: Resources, writing–review and editing. S. Pruitt-Thompson: Data curation, project administration, writing–review and editing. K. Bhushan: Data curation, project administration, writing–review and editing. M.I. Milstein: Data curation, project administration, writing–review and editing. J. Hedglin: Data curation, project administration, writing–review and editing. K.E. Kargus: Resources, writing–review and editing. L.M. Sholl: Formal analysis, writing–review and editing. J. Tsuji: Formal analysis, writing–review and editing. D.M. Hyman: Resources, writing–review and editing. A. Sisk: Data curation, project administration, writing–review and editing. G.I. Shapiro: Resources, project administration, writing–review and editing. H.A. Vargas: Resources, writing–review and editing. J.J. Harding: Resources, writing–review and editing. M.H. Voss: Resources, writing–review and editing. G. Iyer: Resources, formal analysis, writing–review and editing. D.J. Kwiatkowski: Conceptualization, resources, data curation, formal analysis, supervision, methodology, writing–original draft, writing–review and editing.

The funding support, in part, was provided to D.J. Kwiatkowski by Novartis Inc. Patients treated at the Memorial Sloan Kettering Cancer Center were supported in part by Memorial Sloan Kettering Cancer Center Support Grant/Core Grant (P30 CA008748). We thank every person who contributed to this study.

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.