Abstract

   Adenoid cystic carcinoma (ACC) is an aggressive salivary gland
   malignancy with limited treatment options for recurrent or metastatic
   disease. Due to chemotherapy resistance and lack of targeted
   therapeutic approaches, current treatment options for the localized
   disease are limited to surgery and radiation, which fails to prevent
   locoregional recurrences and distant metastases in over 50% of
   patients. Approximately 20% of patients with ACC carry NOTCH-activating
   mutations that are associated with a distinct phenotype, aggressive
   disease, and poor prognosis. Given the role of NOTCH signaling in
   regulating tumor cell behavior, NOTCH inhibitors represent an
   attractive potential therapeutic strategy for this subset of ACC. AL101
   (osugacestat) is a potent γ-secretase inhibitor that prevents
   activation of all four NOTCH receptors. While this investigational new
   drug has demonstrated antineoplastic activity in several preclinical
   cancer models and in patients with advanced solid malignancies, we are
   the first to study the therapeutic benefit of AL101 in ACC. Here, we
   describe the antitumor activity of AL101 using ACC cell lines,
   organoids, and patient-derived xenograft models. Specifically, we find
   that AL101 has potent antitumor effects in in vitro and in vivo models
   of ACC with activating NOTCH1 mutations and constitutively upregulated
   NOTCH signaling pathway, providing a strong rationale for evaluation of
   AL101 in clinical trials for patients with NOTCH-driven
   relapsed/refractory ACC.

   Subject terms: Head and neck cancer, Targeted therapies

Introduction

   Adenoid cystic carcinoma (ACC) is a relatively uncommon secretory gland
   malignancy with a high propensity for perineural invasion, locoregional
   recurrence, and distant metastasis despite curative-intent treatment.
   Due to its insidious infiltrative growth pattern, ACC is often advanced
   by the time of clinical recognition. The therapeutic management for
   patients with locoregional and recurrent/metastatic ACC is limited to
   surgery and radiation, as no systemic agent has been found to be
   effective in improving long-term disease control [[62]1]. The mortality
   rate of ACC remains high, with over 60% of the patients succumbing to
   the disease within 15 years of diagnosis. Therefore, new therapeutic
   approaches for treating ACC are urgently needed.

   Like other malignancies, ACC is thought to arise through genetic and
   epigenetic aberrations that lead to the mis-expression of tumor
   suppressor genes and oncogenes. Molecular and genetic characterization
   of ACC has begun to reveal the most common driver mutations in this
   still incompletely understood cancer [[63]2–[64]5]. One common driver
   mutation in ACC consists of chromosomal rearrangements that produce
   MYB-NFIB (~70% of the tumors) or MYBL1-NFIB fusion genes, which appear
   to have a central role in the genesis of ACC. While representing useful
   markers in the diagnosis of ACC, MYB family members, like other
   transcription factors, remain difficult therapeutic targets, and the
   prognostic and biologic significance of MYB fusion genes are uncertain
   [[65]6]. Other driver mutations in ACC involve genes that encode
   epigenetic regulators (e.g., MLL2, MLL3, EP300, SMARCA2, SMARCC1,
   CREBBP, and KDM6A), pro-growth factors (e.g., FGFR2, PIK3CA, MYC, KRAS,
   BRAF), and DNA damage/checkpoint regulators (ATM, CDKN2A, TP53)
   [[66]2–[67]5, [68]7]. Recurrent copy number losses (e.g., of chromosome
   1p36) and amplifications (e.g., of chromosome 7p14.1 and 14q11.2) have
   also been noted [[69]2–[70]5]. However, the fraction of ACC with
   mutations in these genes that present opportunities for targeted
   therapy is small.

   By contrast, recently emerging data suggest that the NOTCH pathway is a
   tractable, rational therapeutic target in ACC [[71]7, [72]8]. The NOTCH
   gene family contains four paralogs encoding large type 1 transmembrane
   receptor signaling proteins, NOTCH1, 2, 3, and 4. Normal NOTCH
   signaling is initiated by binding the receptor to a ligand (JAG1, JAG2,
   DLL1, or DLL4) on an adjacent cell. This event elicits conformational
   changes in the extracellular domain of NOTCH that makes it susceptible
   to successive cleavages by ADAM10 and γ-secretase. The latter cleavage
   releases the NOTCH intracellular domain (NICD), which migrates to the
   nucleus and forms a transcription complex that upregulates the
   expression of downstream target genes. These upregulated genes, in
   turn, regulate diverse cellular functions in a context-specific
   fashion, many of which have the potential to influence the behavior of
   cancers through both cell autonomous and non-autonomous mechanisms
   [[73]9–[74]13].

   Particular cancers are marked by the presence of NOTCH-activating
   mutations, which either disrupt the extracellular NOTCH negative
   regulatory region (NRR), leading to ligand-independent proteolysis and
   NICD generation, or a C-terminal PEST domain, leading to stabilization
   of NICD [[75]9]. Among these cancers is ACC, in which activating NOTCH
   mutations are found in ~20% of tumors [[76]7, [77]8, [78]14, [79]15].
   According to the largest analysis (n = 1045) conducted across several
   retrospective sequencing datasets [[80]8], these mutations
   predominantly occur in NOTCH1, although mutations in NOTCH2, NOTCH3,
   and NOTCH4 have also been identified [[81]2, [82]5, [83]7, [84]8]. A
   retrospective analysis of three independent cohorts demonstrated that
   ACCs with NOTCH1 activating mutations are associated with a poorer
   median overall survival and progress four times faster than those
   patients without activating mutations [[85]8, [86]15, [87]16]. The
   first retrospective study of 102 ACCs revealed that tumors with NOTCH1
   mutations are associated with solid histology, advanced-stage disease
   at presentation, liver and bone metastases, and shorter relapse-free
   (12.5 vs 33.9 months) and median overall survival (29.6 vs 121.9
   months) compared to tumors with wild-type (WT) NOTCH1 [[88]15]. The
   second retrospective study in 84 patients with recurrent/metastatic ACC
   found that median overall survival was significantly shorter in those
   with NOTCH1-mutated tumors compared to NOTCH1 WT malignancies (55.1 vs
   204.5 months) [[89]8]. Furthermore, among NOTCH1 mutant tumors,
   activating mutations were associated with significantly poorer survival
   (31.1 vs 73.8 months) [[90]8]. These findings were recently supported
   in an independent cohort in which activating NOTCH1 mutations were
   again associated with poorer overall survival (48 vs 195.6 months)
   [[91]16]. Recurrent mutations in genes encoding proteins that regulate
   the activity of the NOTCH transcription complex, such as SPEN and
   FBXW7, have also been identified in ACC genomes, further implicating
   the NOTCH pathway in ACC tumorigenesis [[92]2, [93]7, [94]15].

   Based on the proposed multifaceted pro-oncogenic roles of NOTCH
   signaling in cancer [[95]17–[96]20], several γ-secretase inhibitors
   (GSIs) have been developed [[97]15, [98]21–[99]26] and tested in
   preclinical studies [[100]27–[101]32] and Phase I/II trials in patients
   with advanced solid tumors, including ACC [[102]15, [103]33], either as
   a single agent or in combination with targeted therapeutics or
   chemotherapy [[104]33]. However, the overall response rates to
   NOTCH-targeted therapy have been suboptimal. Factors limiting success
   to date include use of GSIs with suboptimal pharmacokinetic
   (PK)/pharmacodynamic properties, dose-limiting toxicity, and the
   failure to use biomarkers of NOTCH activation as selection criteria for
   clinical trial entry [[105]33–[106]36]. Thus, the potential of NOTCH
   inhibitors such as GSIs has yet to be fully explored.

   AL101 is an investigational small molecule GSI that potently inhibits
   all four NOTCH paralogs and prevents the upregulation of NOTCH target
   genes [[107]22]. While a number of recent preclinical studies have
   shown robust AL101 antitumor activity in several in vivo cancer models
   [[108]22, [109]37], a comprehensive evaluation of its effect in ACC is
   lacking, in part due to the scarcity of experimental model systems for
   ACC.

   Here we demonstrate that AL101 monotherapy inhibits the proliferation
   of NOTCH-activated ACC organoids at nanomolar concentrations.
   Furthermore, AL101 is well tolerated in vivo and has significant
   antitumor effects in ACC patient-derived xenograft (PDX) models with
   NOTCH1 activating mutations, but not in PDX models with WT NOTCH genes.
   Taken together, our results provide a strong foundation for the
   clinical development of AL101 as a targeted monotherapy in patients
   with NOTCH-activated ACC.

Methods

Cell lines and reagents

   The human ACC cell line HACC2A was received from Dr Jacques Nȍr
   (University of Michigan), the UFH2 cell line was received from Dr
   Frederic Kaye (University of Florida), and the ACC52 cell line was
   received from Dr Lurdes Quiemado (University of Oklahoma). Cells were
   monitored for mycoplasma using the MycoDetect kit (Greiner Bio-One).
   HACC2A cells were cultured in DMEM medium (Gibco) supplemented with 10%
   FBS, 200 mM L-glutamine, antibiotic-antimycotic (100X) (Gibco),
   400 ng/ml hydrocortisone, 20 ng/ml epidermal growth factor, 5 µg/ml
   insulin and bovine brain extract (Lonza). UFH2 cells were cultured in
   DMEM+GlutaMAX medium (Gibco) supplemented with 10% FBS and 5000 U/ml
   penicillin-streptomycin (Gibco 15070063). ACC52 cells were cultured in
   RPMI 1640 medium (Corning) supplemented with 10% FBS,
   antibiotic-antimycotic (100X), 20 ng/ml epidermal growth factor,
   400 ng/ml hydrocortisone, and 5 μg/ml insulin. AL101 (batch: 3H66027)
   was obtained from Ayala Pharmaceuticals. All other chemicals used in
   this study were purchased from Sigma and prepared according to the
   manufacturer’s recommendations.

Organoid preparation

   Freshly obtained surgical samples were digested first with a mixture of
   collagenase and dispase, followed by TrypLE (ThermoFisher). After
   passing through a mesh, cells were embedded in Matrigel (Corning) and
   cultured in four different organoid media formulations in parallel.
   These formulations have the same base media containing the growth
   factors EGF, Noggin (Sigma), R-spondin (Sigma), and FGF10
   (ThermoFisher) as well as N-acetylcysteine (ThermoFisher),
   nicotinamide, Y-27632 (Rho kinase inhibitor), A83-01 (TGF-β signaling
   inhibitor), N2, and B27 (all from Sigma). Additional components of
   these media include forskolin (adenylyl cyclase activator), CHIR99021
   (Wnt activator), gastrin I, prostaglandin E2, FGF2, hydrocortisone, and
   heregulin β1 (all from Sigma). The cells that demonstrated the most
   robust growth were further expanded, and fractions of cells were
   harvested for histology, DNA and RNA isolation, or cryopreserved at low
   passage numbers.

IHC staining

   Sections (4-micron) prepared from FFPE PDX tumors were stained for
   NOTCH1 intracellular domain (NICD1) (Cell Signaling, 4147S, 1:200), MYC
   (Abcam, clone Y69, catalog ab32072; 0.56 μg/ml), or Ki67 (BioCare,
   clone SP6, catalog CRM326, 1:100) using a Bond III automated
   immunostainer (Leica). Staining was carried out in the Dana
   Farber/Harvard Cancer Center Specialized Histopathology Core
   Laboratory, which is certified by the College of American Pathologists
   and meets Clinical Laboratory Improvement Amendments standards.
   Staining was developed using the Bond Polymer Refine Detection Kit
   (Leica). Slides were counterstained with hematoxylin and reviewed by a
   board-certified anatomic pathologist (JCA).

Cell viability assays

   For cell lines, relative viability was determined using an Alamar Blue
   assay as outlined by the manufacturer (AbDSerotec). New media
   containing 1/10 volume of Alamar Blue reagent was added to the wells
   and cells were incubated at 37 °C for 1 h. Fluorescence (560 nm
   excitation, 590 nm emission wavelengths) was measured using a
   SpectraMax-Plus384 fluorometer (Sunnyvale). Percent viability was
   determined by comparing DMSO treatment to inhibitor treatment. For
   organoids, ATP measurements (CellTiter-Glo Luminescent Cell Viability
   Assays, Promega) were used to assess proliferation and viability.
   Medium was discarded and CellTiter-Glo 3D reagent was added to each
   well. After incubating 30 min at room temperature on a rotary shaker,
   bioluminescence activity was assessed using a plate luminometer.

Reverse transcription and real-time PCR

   RNA was reverse transcribed to cDNA using Superscript III (Invitrogen)
   and then used as a template for real-time PCR. Gene amplification was
   carried out on a StepOnePlus system (Applied Biosystems) using TaqMan
   Gene Expression Assays (Applied Biosystems). Assay IDs were:
   ACTB-Hs01060665_g1, HEY1-Hs00232618_m1. HEY2-Hs01012057_m1,
   NRARP-Hs04183811_s1, MYC-Hs00153408_m1, and HES5-Hs01387463_g1. All
   reactions were performed in triplicate and relative RNA quantity was
   calculated after normalizing to ACTB expression by the 2^−ΔΔCT method.

Xenograft models

   Early passage PDX tissues were obtained through the Adenoid Cystic
   Carcinoma Research Foundation. All animal procedures were performed at
   XenoSTART (South Texas Accelerated Research Therapeutics, San Antonio)
   following Institutional Animal Care and Use Committee protocols.
   Fragments of tumor (~70 mm^3) were implanted subcutaneously into the
   flanks of 6–12-week-old female nu/nu athymic nude mice (The Jackson
   Laboratories or Charles River Laboratories). Upon reaching 150–300 mm^3
   tumor volume, mice were randomized to either treatment (n = 5) or
   vehicle (n = 5–10) groups using blinded block randomization and
   therapeutic dosing was implemented. Tumor dimensions were measured
   using digital calipers blinded to the treatment group and tumor volume
   was calculated using the formula: TV = width^2 × length × 0.52. Tumors
   were harvested at termination, weighted, and used for histology,
   immunohistochemical staining, and RNA sequencing (RNA-seq) analysis
   (n = 3–5 per arm). Percent mean tumor growth inhibition (%TGI) induced
   by AL101 was calculated relative to the untreated control group using
   the formula: %TGI = 1 – (AL101 final – AL101 baseline) / (Control
   final – Control baseline).

NOTCH luciferase reporter assay

   Full-length (FL) human NOTCH1 (FL-NOTCH1), ΔECD-NOTCH1, ACCx9-I1680N,
   and ACCx11-S1723ins28 transcripts were synthesized by GenScript and
   ligated into the pcDNA3.1 (+FLAG) expression vector. Constructs were
   transiently transfected into U2OS cells using Lipofectamine 2000
   reagent (Invitrogen, #11668019) and assessed for their ability to
   activate a NOTCH-sensitive luciferase reporter gene [[110]38, [111]39].
   U2OS cells were used because of their transfectability and low basal
   NOTCH activity. Briefly, cells were co-transfected (in five biological
   replicates) with 10 ng of pcDNA3.1 expression construct, a
   NOTCH-sensitive firefly luciferase reporter gene (TP1-CSLx12-FF), and
   an internal control Renilla luciferase plasmid (pRL-TK, Promega).
   Normalized firefly luciferase activities were measured in whole cell
   extracts prepared 48 h after transfection using the Dual-Luciferase kit
   (Promega #E1960) and an Infinite M200 luminometer (Tecan).

RNA sequencing

   RNA-seq was performed on an Illumina NovaSeq-6000 instrument. Adapters
   were trimmed with Cutadapt [[112]40]. For PDXs, mouse reads were
   filtered out using an approach described by Callari et al. [[113]41].
   Human reads were aligned to human reference genome GRCh37/hg19 using
   STAR and STAR-fusion [[114]42, [115]43]. Gene expression levels were
   calculated using featureCounts [[116]44] and gene expression levels
   were normalized using DESeq2 [[117]45]. Differentially expressed (DE)
   genes were detected using DESeq2 according to the following parameters:
   (i) average gene expression >50 normalized reads, (ii)
   log2(fold-change) >1 or log2(fold-change) <−1, (iii) false discovery
   rate (FDR) < 0.05.

Pathway enrichment analysis

   Pathway enrichment analysis was performed using minimal hyper-geometric
   statistics [[118]46], based on an approach similar to that described by
   Eden et al. [[119]47]. MSigDB c2 (curated gene sets: Kyoto Encyclopedia
   of Genes and Genomes (KEGG)/Reactome pathways) [[120]48] were used as
   pathway references. Multiple hypothesis correction of the p values was