Discovery and Characterization of Recurrent, Targetable ALK Fusions in Leiomyosarcoma

This article is featured in Highlights of This Issue, p. 667

Soft-tissue sarcomas (STS) are a heterogeneous group of rare tumors of mesenchymal origin. Despite multimodal treatment including surgery, chemotherapy, and radiation, STS frequently portend a poor prognosis (1–3). Discovery and validation of oncogenic drivers and cognate-targeted therapies has the potential to dramatically impact patient outcomes by providing more effective, less morbid treatment.

Constitutively activated kinase fusions arising from somatic chromosomal rearrangements are known drivers of malignant transformation (4). Approximately 20% of sarcomas have well-known translocations, such as the SS18-SSX fusion in synovial sarcoma. However, few rearrangements found in sarcomas involve a kinase partner, and thus are not targetable with currently available small-molecule inhibitors. The exceptions are in rare, locally aggressive tumors, including dermatofibrosarcoma protuberans (COL1A1-PDGFB) and tenosynovial giant cell tumors (COL6A3-CSF1) and infantile fibrosarcoma (ETV6-NTRK3; refs. 5–7). Targetable nonfusion kinase alterations have been identified in a handful of sarcoma subtypes, most notably activation of cKIT and PDGFR receptors in gastrointestinal stromal tumors (GIST), where treatment with kinase inhibitors has dramatically changed prognosis (8).

Leiomyosarcoma is the second most common sarcoma, with approximately 2,000 newly diagnosed patients per year (9–11). Leiomyosarcomas originate from smooth muscle, most often of the uterus, retroperitoneum, or blood vessels, and leiomyosarcomas frequently develop resistance to standard cytotoxic chemotherapy (12). The genomic landscape of leiomyosarcoma tends to be complex, and previously there was no indication of actionable or recurrent gene rearrangements. The Cancer Genome Atlas (TCGA) Research Network's comprehensive characterization of soft-tissue sarcomas included 80 leiomyosarcoma specimens; no oncogenic kinase fusions were reported (13). However, multiple targetable kinase fusions were reported previously from a set of 967 sarcomas sequenced by Foundation Medicine as a part of clinical care (14, 15). In that study, we identified 5 leiomyosarcomas with ALK fusions, leading us to hypothesize that ALK fusions may be an underappreciated and therapeutically targetable occurrence within a clinically meaningful subset of patients with leiomyosarcoma.

Anaplastic lymphoma kinase (ALK), a receptor tyrosine kinase, is primarily expressed in the developing central nervous system (CNS) and has weak to no expression in normal adult human tissues (16). Chromosomal rearrangements of the ALK gene generate chimeric ALK fusion proteins wherein the rearrangement retains the ALK kinase domain sequence with various in-frame 5′ partners (4). ALK rearrangements are established oncogenic drivers in a subset of non–small cell lung cancer (NSCLC; ref. 17), but have also been reported in several other malignancies, including in leukemia, B-cell lymphoma, inflammatory myofibroblastic tumor (IMT), and anaplastic large cell lymphoma (ALCL; refs. 4, 18–23).

The discovery of EML4-ALK in a subset of patients with NSCLC propelled the development of small-molecule ALK inhibitors. Crizotinib, the first FDA-approved ALK inhibitor, was rapidly implemented as standard of care for patients with ALK fusion–positive lung adenocarcinoma (24–28). Several second- and third-generation ALK inhibitors are now in clinical trials, including lorlatinib and entrectinib (29, 30), and three additional ALK inhibitors were recently FDA approved: ceritinib (Novartis, April 2014), alectinib (Hoffman-La Roche, December 2015), and brigatinib (Ariad, April 2017). ALK tyrosine kinase inhibitors are expected to offer significant therapeutic benefit to patients with ALK-rearranged malignancies beyond lung cancer.

Following the identification of 5 of 223 leiomyosarcomas with ALK fusions in the Foundation Medicine dataset (14), we explored the incidence, diversity and functional impact of ALK fusions in leiomyosarcoma in two additional independent cohorts, namely TCGA and a tissue microarray (TMA). We report the discovery and characterization of recurrent, targetable ALK rearrangements identified in these cohorts. We go on to demonstrate efficacy of clinically viable ALK tyrosine kinase inhibitors in multiple cell-based model systems and highlight that these fusions are oncogenic drivers in a lineage-specific murine model of leiomyosarcoma. Taken together, these data have immediate translational relevance to potentially impact diagnostic approaches, treatment, and prognosis in this aggressive cancer subtype.

The TCGA RNA-seq data were downloaded from the dbGaP website, under phs000178.v8.p7 (31). We applied four different software packages for the identification of ALK rearrangements in RNA-Seq data of selected leiomyosarcoma TCGA cases: STAR-Fusion version 0.1.1 (32), ChimeraScan version 0.4.5 (33), deFuse version 0.6.2 (34), and FusionCatcher (35) version 0.99.5a. The fusion calling algorithms were run on the RNA-Seq fastq files with paired-end reads generated by Illumina HiSeq2000 instrument with the default settings.

We employed a previously described TMA, TA-381, which includes 126 unique leiomyosarcoma donor samples (36). The FISH assay was performed to examine ALK rearrangement (Biocare Medical catalog no. 902-7000-082115) using 4-μm formalin-fixed paraffin-embedded (FFPE) sections. As per the manufacturer, the 3′ ALK probe was labeled in red (Ex 593 nm; Em 618 nm) while the 5′ ALK probe was labeled in green (Ex 498 nm; Em 522 nm). Slides were deparaffinized in xylene ×3 for 10 minutes, dehydrated twice with 100% ethanol, air dried for 10 minutes, and then pretreated in 10 mmol/L citric acid pH 6.0 at 80°C for 45 minutes. Samples on slides were digested for 60 minutes in pepsin (at 75,000 U, Sigma catalog no. P6887) at 37°C. Fluorescence-labeled probes and slides were codenatured at 75°C for 7 minutes and hybridized at 37°C for 16–18 hours in a humidified chamber. Posthybridization washes were performed using 2 × SSC/0.3% NP-40 at 72°C for 5 minutes. Slides were dehydrated and air-dried in the dark and counterstained with DAPI (catalog no. P36935 Invitrogen/Life Technologies), and followed by imaging. Probe deemed evaluable in 97 cases. Signals were scored by a cytogeneticist (M. Debiec-Rychter) by evaluating at least 50 tumor cell nuclei per case. A split signal was defined by 5′ and 3′ signals being observed at a distance greater than a single signal width, and signals separated by less than this were regarded as fused signals. Tumor cells showing split signals or isolated 3′ (red labeled) signals were considered to have rearrangements of ALK. We interpreted a result as FISH-positive if >20% of tumor cells showed gene rearrangement.

Small sections of FFPE tumor tissue were excised, placed in xylene, and vortexed vigorously. Tissue homogenate was fractionated by centrifugation; the supernatant was discarded and the pellet was washed in ethanol by vortexing and recentrifugation, followed by air-drying of the ethanol-treated pellet. An RNA extraction kit (RNAeasy FFPE kit, Qiagen) was used to purify RNA from tissue and amount of RNA was quantified by measuring absorbance using the NanoDrop spectrophotometer. cDNA was made using VILO kit (Thermo Fisher Scientific). Primers used for amplification of ALK using PCR were TGATGATCAGGGCTTCCATG and CATCACCCTCATTCGGGGT taken from 20th exon. Primers for GAPDH were AGGGCTGCTTTTAACTCTGGT and CCCCACTTGATTTTGGAGGGA. RT-PCR using VILO reagents were conducted on C1000 Touch Thermal Cycler (Bio-Rad) with a cycle of 94°C for 15 seconds, followed by 45 cycles of 94°C for 15 seconds, 56°C for 30 seconds, and 68°C for 30 seconds. PCR products were resolved on a 2% agarose gel and imaged using Gel DOC^tm XR+ (Bio-Rad).

ACTG2-ALK and KANK2-ALK cDNA were generated using GenScript gene synthesis service in pUC57 vector, and subsequently subcloned into the pMSCV-IRES-GFP retroviral vector using In-Fusion HD cloning (Clontech) following manufacturer's protocol. pMX-p53DD was a gift from Dr. Shinya Yamanaka, Kyoto University, Japan (Addgene plasmid # 22729; ref. 37), and the p53DD insert from this vector was subcloned into a pMSCV-IRES-mRFP retroviral vector using In-Fusion HD Cloning Kit (Clontech) to enable enrichment of dual GFP (ALK fusion) and mRFP (p53DD)-positive cells populations.

Where indicated, retroviral particles were produced by transient transfection of the ecotropic viral packaging cell line, Platinum-E (Cell Biolabs) following manufacturer's protocol. Culture supernatant containing retroviral particles was collected 48 hours after transfection, and immediately used for transduction of recipient cells (Ba/F3, NIH3T3, or mSMCs).

Ba/F3 cells (ATCC) were cultured in RPMI1640 medium (Thermo Fisher Scientific) with 10% (vol/vol) FBS, l-glutamine, penicillin/streptomycin, and 15% (vol/vol) conditioned media produced from the WEHI-3 myelomonocytic IL3-secreting cells. Ba/F3 cell lines were maintained at densities of 0.5 to 1.5 × 10⁶/mL and infected with retrovirus-encoding human KANK2-ALK and ACTG2-ALK fusion genes. Stable cell lines were generated after FACS for GFP. Cell number and viability were counted every 2–3 days. NIH 3T3 cells (ATCC) were cultured in DMEM-high glucose medium (Thermo Fisher Scientific) supplemented with 10% FCS, l-glutamine, and penicillin/streptomycin. Early-passage NIH 3T3 cells were transduced with KANK2-ALK and sorted for GFP expression as described for Ba/F3 cells. Primary uterine mouse smooth muscle cells (mSMC) from C57BL/6 mice were purchased from Cell Biologics, and cultured in growth media until 75% confluent. Normal mSMCs went into quiescence after 2 weeks, but colonies of transduced dnp53 mSMCs as detected by mRFP expression were clonally expanded; dnp53 expression induced mSMC immortalization without change to morphology. Once established, the mSMC dnp53 (p53DD) cell line was cultured in DMEM-high glucose medium supplemented with 10% FCS, l-glutamine, and penicillin/streptomycin. For bioluminescent imaging, stable mSMC dnp53 cells were transduced with firefly luciferase lentiviral particles (Cellomics Technology) and selected with blasticidin. In a final step, the mSMC dnp53 plus firefly luciferase cells were transduced with KANK2-ALK retroviral particles as described above.

For dose–response cell viability assays, Ba/F3 cell lines (1 × 10³ cells) were distributed in 384-well plates that had escalating concentrations of inhibitors (crizotinib, ceritinib, alectinib, lorlatinib, entrectinib, brigatinib) ranging from 1 pmol/L to 1 μmol/L, and incubated for 72 hours. Cell viability was measured using a methanethiosulfonate (MTS)-based assay (CellTiter96 Aqueous One Solution; Promega) and absorbance (490 nm) was read after 1 hour using the BioTek Synergy 2 plate reader. To facilitate comparison between experiments, the raw MTS absorbance of vehicle (0.1% DMSO)-treated cells was set to 1, and absorbance from inhibitor-treated wells was normalized to the vehicle-treated value. Each experiment had a minimum of three replicate wells per condition and the average and SEM was plotted for curve fit analysis. Assay was repeated on two additional days; results are from one representative experiment. Data were normalized using Microsoft Excel and further nonlinear regression curve fit analysis of the normalized data for determination of IC₅₀ values was performed using GraphPad Prism software.

NIH3T3 or mSMC cells stably expressing KANK2-ALK were treated for 3 hours with vehicle, lorlatinib, crizotinib, and entrectinib at doses as indicated in Fig. 2. Immediately after treatment, the cells were washed once in Dulbecco PBS and collected in 100 μL of cell lysis buffer [20 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 1% Triton, 2.5 mmol/L sodium pyrophosphate, 1 mmol/L beta-glycerophosphate, 1 mmol/L sodium orthovanadate, 1 μg/mL leupeptin and 1 mmol/L phenylmethylsulfonyl fluoride]. Protein content was assessed by bicinchoninic acid assay (Thermo Fisher Scientific), and 40 μg of protein lysate was evaluated by gel electrophoresis and immunoblotting with following antibodies: total ALK (3633S, Cell Signaling Technology), phospho-ALK (3983S, Cell Signaling Technology), phospho-ERK1/2 (9101, Cell Signaling Technology), and GAPDH (AM4300, Thermo Fisher Scientific).

All animal protocols were executed in compliance with institutional guidelines and regulations, and after approval from the OHSU Institutional Animal Care and Use Committee (protocol IP00000031). Eight-week-old female nude mice (The Jackson Laboratory) were injected with dnp53 mSMC or KANK2-ALK mSMC (1 × 10⁶ cells in 50 μL PBS) subcutaneously. In vivo luciferase imaging was performed 10 minutes after mice were injected (intraperitoneal) with luciferin (150 mg/kg, 15 mg/mL in sterile PBS, XenoLight D-Luciferin, PerkinElmer), using the IVIS system (IVIS Spectrum, Caliper Life Sciences). Mice were monitored for tumor engraftment, and once palpable, tumor growth was monitored using digital calipers (Fisher Scientific). Once tumor size reached 0.2 cc³, mice were randomly assigned to one of the following treatment groups: vehicle alone (n = 7, 0.5% methyl cellulose/0.5% Tween-80), lorlatinib (n = 6, 3 mg/kg, provided by Pfizer via MTA and then prepared in 10% ethanol, 40% PEG200), or crizotinib (n = 6, 100 mg/kg in 0.5% methyl cellulose/0.5% Tween-80; purchased from LC Labs). Each group was dosed via oral gavage once daily 5 days per week for 4 weeks. Mice were weighed and tumor measured thrice weekly. Once tumors reached 2 cc³, animals were euthanized and tumors were harvested for biochemical (flash frozen) and IHC (10% zinc formalin) analyses. For on-target evaluation of lorlatinib on intratumoral ALK, small pieces of flash frozen, treated tumors were placed in liquid nitrogen and pulverized using a mortar and pestle prior to protein extraction with cell lysis buffer. Forty micrograms of tumor lysate was evaluated by immunoblotting as described above.

Where indicated, the Student t test (Microsoft Excel or GraphPad Prism) was used to determine statistical significance and comparisons; P values less than 0.05 were deemed significant. Asterisks in figures and the description of asterisks in figure legends indicate level of statistical significance.

Kinase fusions are being reported at a low incidence across a wide variety of malignancies, leading to profound benefit for those patients (4, 6, 29, 38). Given the clear clinical need for improved therapeutic options, we sought to identify clinically relevant genomic alterations in sarcomas. We previously interrogated next-generation sequencing data performed as part of clinical care (14), and discovered ALK rearrangements in 5 of 223 leiomyosarcomas (Table 1, “FM-LMS” samples). The gene partner for FM-LMS5 was not identified; the Foundation Medicine raw sequencing data are proprietary but the Foundation Medicine analysis algorithms considered this fusion clinically reportable. Concurrent loss of CDKN2A/B and/or alteration of TP53 was present in all five samples.

The frequency of ALK rearrangements specifically in leiomyosarcomas in the Foundation Medicine cohort prompted us to interrogate The Cancer Genome Atlas (TCGA) Network's sarcoma genomic dataset for evidence of ALK fusions. On the basis of the bioinformatics algorithm developed by Shah and colleagues (39), we assessed the publicly available TCGA sarcoma DNA copy number data (provisional dataset, available through cBioPortal; ref. 40) for 5′/3′ imbalances in the ALK gene. Three of 57 LMS samples demonstrated DNA copy number alterations within the ALK gene, suggestive of an unbalanced ALK rearrangement. Visualization of these data using the Integrative Genomics Viewer (41) indicated loss of a chromosomal segment just upstream of the ALK kinase domain. All three specimens were FNCLCC grade 2 uterine leiomyosarcoma (13).

Subsequent detailed analysis of raw RNA-sequencing data for two samples, TCGA-FX-A48G and TCGA-IW-A3M6, revealed novel fusion transcripts (Table 1; Fig. 1A and B; Supplementary Fig. S1). One of the three specimens showing genomic imbalance in the ALK gene (TCGA-IS-A3K7) did not have adequate read depth by RNA-Seq to validate the rearrangement. The ACTG2-ALK fusion found in TCGA-IW-A3M6 and the KANK2-ALK fusion identified in TCGA-FX-A48G are transcripts with unknown oncogenic potential, and KANK2-ALK has not previously been reported. The ACTG2-ALK fusion found in TCGA-IW-A3M6 was identified by DeFuse, FusionCatcher, and ChimeraScan, and the KANK2-ALK fusion in TCGA-FX-A48G was identified by DeFuse and ChimeraScan (Supplementary Table S1; Supplementary Figs. S2 and S3). Notably, the ACTG2-ALK discovered in the TCGA dataset was also independently identified in a distinct clinical sequencing sample, FM-LMS-3, from a separate cohort (Table 1). ACTG2 (2p13.1) and ALK (2p23.2-p23.1) are located on chromosome 2p in opposite orientations, suggesting that a cryptic paracentric inversion is responsible for this rearrangement.

To validate ALK fusions in leiomyosarcoma and ascertain incidence, we examined ALK rearrangements in an independent leiomyosarcoma patient cohort. We assessed a leiomyosarcoma tissue microarray (TA-381) using FISH; all samples in the TMA underwent expert pathologic review prior to construction (36). Among 97 leiomyosarcoma cases evaluable by FISH, two samples exhibited ALK rearrangement as detected by ALK break-apart FISH probes (TA-381-6806 and TA-381-6871, Fig. 1C). Both of the ALK-rearranged cases expressed the ALK kinase domain as detected by reverse-transcriptase PCR (RT-PCR; Fig. 1D) while samples with normal FISH signal expressed detectable levels of ALK (representative example 6021 in Fig. 1D). We were unable to characterize the specific 5′ ALK fusion partner in the ALK-rearranged TMA samples, TA-381-6806 and TA-381-6871, due to insufficient tissue availability.

Because KANK2-ALK was previously uncharacterized, we assessed oncogenic potential. Three independent model systems were employed to test whether: (i) KANK2-ALK functions as a dominant, transforming oncogene in the well-established Ba/F3 model system; (ii) clinically-relevant ALK inhibitors effectively inhibit ALK catalytic activity and downstream signaling in KANK2-ALK expressing NIH3T3 murine fibroblasts; and (iii) KANK2-ALK drives tumor growth in an in vivo, lineage-specific murine tumor model.

Ba/F3, a pro-B murine cell line, is an established model system for investigating transformative potential of oncogenic kinase variants (42). Expression of constitutively activated tyrosine kinases, including kinase fusion oncogenes, confers IL3 independence to permit limitless replicative growth, thus transforming Ba/F3 cells. Stable expression of KANK2-ALK provided sufficient signaling drive for rapid and sustained IL3-independent Ba/F3 outgrowth, while cells expressing empty vector (pMSCV-PIG) were unable to grow without IL3 (Fig. 2A; Supplementary Fig. S4). We determined the cell-based sensitivity of onco-addicted Ba/F3 KANK2-ALK cells to multiple ALK inhibitors, including crizotinib, ceritinib, alectinib, lorlatinib, entrectinib, and brigatinib, using dose–response cell viability assays (Fig. 2B and C; Supplementary Fig. S5). Growth and viability of Ba/F3 KANK2-ALK cell lines were inhibited to varying degrees by all inhibitors tested. Lorlatinib was the most potent inhibitor, exhibiting cell-based 50% inhibitory concentrations (IC₅₀) of 0.14 nmol/L. The other inhibitors tested were also effective against KANK2-ALK: crizotinib IC₅₀ = 11.3 nmol/L; ceritinib IC₅₀ = 10.8 nmol/L; brigatinib IC₅₀ = 7.2 nmol/L; entrectinib IC₅₀ = 4.2 nmol/L; alectinib IC₅₀ = 602 nmol/L. These data demonstrate that multiple ALK inhibitors may be effective against KANK2-ALK fusion–driven cells.

The propensity of transformed cells to proliferate and survive independent of a solid surface is one of the hallmark properties of neoplastic transformation (43, 44). The soft-agar colony formation assay is an established method for assessing oncogenic activity and malignant transformation using murine NIH3T3 fibroblasts. We engineered a NIH3T3 KANK2-ALK cell line using retrovirus, and performed the soft-agar assay to evaluate colony formation as compared with NIH3T3 parental cells. KANK2-ALK increased the ability of NIH3T3 cells to form colonies in soft agar. Furthermore, colony formation induced by KANK2-ALK was suppressed in a dose-dependent manner with lorlatinib and entrectinib (Fig. 2D and E). On-target inhibition of KANK2-ALK catalytic activity was confirmed biochemically by immunoblotting of lysates generated from stable NIH3T3 KANK2-ALK cell with phospho-ALK antibody (Fig. 2F). Robust inhibition of ALK autophosphorylation is observed with entrectinib and lorlatinib as compared with weaker inhibition with crizotinib.

KANK2-ALK functions as a dominant oncogene in both Ba/F3 and NIH3T3 model systems of cell transformation (Fig. 2). Because ALK fusions have not been previously interrogated as drivers of leiomyosarcoma, we wanted to determine whether ectopic expression of an ALK fusion protein is sufficient to induce neoplastic transformation and tumorigenicity in smooth muscle cells, the presumptive cell of origin for leiomyosarcoma (45).

Ba/F3 and NIH3T3 cells are immortalized cell lines, while primary nonneoplastic cells such as mSMCs frequently undergo oncogene-induced senescence in culture. Extensive previous work has demonstrated that deficiency in p53 is critical for neoplastic transformation of primary cells in model systems (44, 46–48). Therefore, we first generated a stable murine smooth muscle cells (mSMC) cell line that expressed a transdominant negative p53 construct (p53DD; refs. 37, 49) followed by serial transduction and selection of KANK2-ALK.

KANK2-ALK is expressed in the transformed mSMCs, and is sensitive to inhibition by lorlatinib, entrectinib, and crizotinib, with lorlatinib being the most potent inhibitor resulting in >50% phospho-inhibition at 10 nmol/L (Fig. 3A). In mSMCs, KANK2-ALK kinase activity is coupled to downstream effector signaling via the canonical RAS/MAPK pathway, as shown by ALK tyrosine kinase inhibitor suppression of ERK phosphorylation in concordance with ALK phospho-inhibition (Fig. 3A). These data suggest that mSMC KANK2-ALK is an adequate lineage-specific model system for determining tumorigenicity in vivo.

To determine the tumor-forming potential of mSMC KANK2-ALK cells, we performed subcutaneous flank implantation of these engineered mSMCs in Nu/Nu mice. For longitudinal monitoring of cell engraftment and tumor growth using noninvasive bioluminescent imaging, the mSMC p53DD (control) and mSMC p53DD, KANK2-ALK cells were lentivirally engineered and selected to stably express Renilla luciferase. Luciferase-positive tumor cell engraftment and nascent tumor formation is apparent on day 3 after injection of KANK2-ALK cells but not of mSMC p53DD cells (Fig. 3B). Tumor volume was measurable using calipers by day 14. Nineteen days after implantation, the mSMC KANK2-ALK tumors reached a volume of 0.2 cm³. At this time, the mice were randomized into three cohorts and treated with vehicle, crizotinib, or lorlatinib by oral gavage. Lorlatinib resulted in near immediate tumor regression (Fig. 3C). Treatment continued for 4 weeks or until tumor volume exceeded 2 cm³, at which point animals were humanely sacrificed. Median survival of tumor-bearing mice improved modestly with crizotinib treatment (36 vs. 47 days), but was significantly extended with lorlatinib treatment (median OS not achieved, P < 0.001; Fig. 3D). Mice in the lorlatinib cohort were followed over time and developed measurable tumors 10–15 days after drug discontinuation (Fig. 3C).

To assess on-target inhibition of KANK2-ALK catalytic activity, mice were orally treated with lorlatinib for five days and tissue was harvested following last treatment for immunoblotting with phospho- and total ALK antibodies (Fig. 3E). Lorlatinib treatment abrogated phospho-ALK as hypothesized. Loss of phospho-ALK is concomitant with decrease of total ALK level in lorlatinib versus untreated tumors that is presumably via protein destabilization and degradation as has been previously reported for EML4-ALK variants (50). Both phospho- and total ALK levels are restored to control levels at 48 hours after lorlatinib treatment was stopped (Fig. 3E, lanes 5 and 6). IHC of lorlatinib-treated tumors confirms downstream ALK effector signaling as evidenced by diminished phospho-S6 staining concurrent with areas of necrosis, and induction of apoptosis as seen by hematoxylin and eosin (H&E) and cleaved caspase-3 staining (Fig. 3F).

Over the last decade, implementation of advanced next-generation sequencing (NGS) technologies in molecular pathology and clinical diagnostics has enabled the discovery of novel oncogenic drivers, permitted tumor reclassification and, in some cases, guided treatment decisions. Targeted therapy directed at a dominant oncogenic driver has dramatically altered survival outcomes in subsets of patients from diverse malignancies including chronic myeloid leukemia, NSCLC, and GIST (26, 51, 52). However, not all patients with cancer have benefitted from these advancements yet, particularly patients with rare cancer subtypes.

Soft-tissue sarcomas are uncommon cancers with a high degree of histologic and genomic heterogeneity, which has stymied the development of effective new treatment strategies. The recent characterization of these tumors by TCGA offers new opportunities for unraveling disease biology and uncovering pharmacologically targetable vulnerabilities (13). By interrogating these data as well as similar NGS data from Foundation Medicine (14, 15), we have found that a subset of leiomyosarcoma tumors harbor ALK fusions that retain the ALK kinase domain.

The KANK2-ALK and ACTG2-ALK fusions reported here contain 5′ fusion partner genes that have not previously been validated in cancer. Notably, these ALK fusions have breakpoints within introns 15 (KANK2-ALK) or 17 (ACTG2-ALK), as opposed to the previously reported and more common intron 19 breakpoint fusions such as in case of EML4-ALK (38). We have shown that KANK2-ALK has robust oncogenic potential and is sensitive to ALK inhibitors—similar to ALK fusions with intron 19 breakpoint.

Because ACTG2 and ALK are located in opposite orientations on chromosome 2p, the formation of ACTG2-ALK is likely a cryptic paracentric inversion mechanism, which is distinct from EML4-ALK formation and not detectable by karyotyping or comparative genomic hybridization (CGH). FISH results may also be misleading in some cases. In our study, the FISH-positive leiomyosarcoma cells (Fig. 1C) demonstrate loss of the 5′ region of ALK. While this may represent a false-positive, this pattern is consistent with the presence of an unbalanced translocation, a common mechanism in lymphomas and IMTs (53). Alternatively, loss of the 5′ signal is also seen in fusions generated from intrachromosomal deletion, such as those known to occur in the ALK paralog, ROS1 (54). Future studies in which sufficient material is available could compare results of FISH with sequencing data to better establish the rate of FISH false-positive results.

The difficulty in fusion gene discovery extends beyond paracentric inversions, unbalanced translocations, and intrachromosomal deletions. Even with NGS data, the bioinformatic analytic processes routinely applied for mutation or copy number analyses are often inadequate. The ongoing development of advanced algorithms for fusion detection [e.g., STAR-Fusion (32), ChimeraScan (33), deFuse (34), FusionCatcher (35)] is resulting in higher sensitivity and specificity in detecting and calling fusion genes.

In cancer cells, various structural rearrangements may exist at the genomic level, but whether the chimeric fusion protein is expressed or not is largely dictated by the transcriptional activity at the promoter of the 5′ fusion partner gene. Thus, we can posit that the expression of the 5′ partner gene is a form of selection pressure dictating which chromosomal translocations are present in diverse cancer cell lineages. Given this hypothesis, it is interesting to note that ACTG2 codes for enteric smooth muscle actin gamma 2, a protein that is highly expressed in smooth muscle cells within the gastrointestinal and genitourinary system (55). Therefore, it is rational that the ACTG2 promoter drives expression of an oncogenic ACTG2-ALK fusion gene in a smooth muscle cell, ultimately resulting in a leiomyosarcoma. In fact, ACTG2 has been used as an IHC marker to identify subtype 1 leiomyosarcomas, which have a gene expression profile more similar to normal smooth muscle than other leiomyosarcoma subtypes (36). Similarly, KANK2 encodes an ankyrin domain–containing cytoskeletal regulatory protein that is robustly expressed in smooth muscle and uterus, again suggestive of selection of these fusions in smooth muscle tumors due to active promoters driving expression of the ALK fusion (55).

Having received FDA approval in 2011, crizotinib is a well-established the first-generation ALK inhibitor in clinical use. However, the pharmacologic pipeline for targeting ALK is expanding, with four drugs now FDA-approved and several other agents in various stages of development (24–30). Lorlatinib is currently in phase III clinical trials, most notably being evaluated against crizotinib for first-line treatment of patients with ALK-positive NSCLC (NCT03052608). Given the exquisite sub- to low nanomolar in vitro sensitivity of Ba/F3 ACTG2-ALK and KANK2-ALK cells to lorlatinib, we compared efficacy of oral monotherapy with crizotinib and lorlatinib in the murine model of KANK2-ALK leiomyosarcoma. Crizotinib offered only modest benefit both in terms of survival and inhibition of tumor growth, while lorlatinib induced statistically significant tumor regression and prolonged survival. This effect is despite the use of relatively high doses of crizotinib (100 mg/kg daily) and moderate doses of lorlatinib (3 mg/kg). Notably, in vitro treatment of NIH3T3 and mSMC KANK2-ALK cells with crizotinib showed similar weak inhibitory activity. A similar trend was previously observed in a ROS1 fusion–driven murine cholangiocarcinoma model, where in vivo efficacy of oral crizotinib was modest as compared with type II ROS1 inhibitor, foretinib (56). Lorlatinib was specifically engineered to exhibit high potency and is effective against multiple cancer types in vitro and in vivo at lower doses than crizotinib (57–59). In addition, we postulate that structural differences in the KANK2-ALK fusion protein may lead to improved binding of lorlatinib as compared with crizotinib, as it retains Exon 16 and therefore a considerable amount of the juxtamembrane region preceding the kinase domain, which may influence binding pocket conformation. Future studies using molecular dynamic simulation may explain the observed differences in inhibitor efficacy.

The presence of an ALK rearrangement in a soft-tissue tumor is most often associated with inflammatory myofibroblastic tumor (IMT). Although both IMT and leiomyosarcoma are mesenchymal soft-tissue neoplasms, each have with distinct histology, epidemiology and metastatic potential. In contrast to leiomyosarcoma, IMT occurs primarily in children and young adults and rarely metastasizes. IMT is a distinctive tumor with an abundance of inflammatory infiltrate and fibroblastic spindle cells and is easily distinguishable from leiomyosarcoma by an experienced sarcoma pathologist. Up to 70% of IMTs contain clonal rearrangements involving ALK (12). The tumors harboring ACTG2-ALK and KANK2-ALK included in our analyses were TCGA specimens; as such, these samples were evaluated by multiple sarcoma pathologists and confirmed to be leiomyosarcoma. Thus, the presence of an ALK rearrangement does not change the histologic diagnosis arrived at by morphologic and IHC assessment. In addition, these ALK fusions have not been previously reported in IMT.

The Cancer Genome Atlas sarcoma dataset includes next-generation sequencing of DNA and RNA as well as additional in-depth molecular evaluations that would not typically be obtained as part of clinical care. Even with the depth of these raw data and detailed, expert analyses, ACTG2-ALK and KANK2-ALK fusions were not previously identified. However, systematic, hypothesis-based interrogation of existing datasets, such as the approach employed here, may yield clinically meaningful discoveries of targetable oncogenes in sarcoma.

With improving ability to detect rearrangements and multiple highly effective and safe treatment options for ALK-positive disease, we believe advanced molecular diagnostics should be employed more often in the sarcoma patient population. Cumulatively, our data suggest that ALK and potentially other targetable kinase fusions may be present in subsets of soft-tissue sarcoma patients, and their discovery may propel new therapeutic strategies in this treatment-refractory cancer.

No potential conflicts of interest were disclosed.

Conception and design: L.E. Davis, K. Nusser, M.A. Davare

Development of methodology: L.E. Davis, K. Nusser, J. Przybyl, N.E. Hofmann, M. Debiec-Rychter, M.A. Davare

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.E. Davis, K. Nusser, J. Przybyl, N.E. Hofmann, M. van de Rijn, M.A. Davare

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.E. Davis, K. Nusser, J. Przybyl, S. Varma, S. Vennam, M. Debiec-Rychter, M. van de Rijn, M.A. Davare

Writing, review, and/or revision of the manuscript: L.E. Davis, K. Nusser, J. Przybyl, M. Debiec-Rychter, M. van de Rijn, M.A. Davare

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.E. Davis, K. Nusser, J. Pittsenbarger, N.E. Hofmann, M.A. Davare

Study supervision: L.E. Davis, M.A. Davare

This work was supported by the Knight Cancer Institute Pilot Grant (intramural).

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.