Implication of ZNF217 in Accelerating Tumor Development and Therapeutically Targeting ZNF217-Induced PI3K–AKT Signaling for the Treatment of Metastatic Osteosarcoma

Osteosarcoma is a heterogeneous, rare malignancy of the bone commonly arising in children and adolescents (1). Surgical resection and combinatorial chemotherapy are beneficial to approximately 70% of localized cases, but patients with advanced metastatic and/or relapsed disease continue to have poor survival outcomes (1, 2). Despite remarkable progress in advancing our knowledge of osteosarcoma, disease recurrence (3), and chemotherapeutic resistance (4, 5) continue to be major roadblocks to curative successes. This significantly underlines the need for new, meaningful therapeutic targets. For these reasons, we performed a Sleeping Beauty (SB) transposon-based forward genetic screen for osteosarcoma which identified >250 previously known and unknown drivers of osteosarcoma development and metastasis (6). In particular, zinc finger protein transcription factor (TF) ZNF217 (murine Zfp217) was identified as a candidate osteosarcoma driver oncogene in a subset of our primary murine osteosarcoma samples (6).

ZNF217 is a member of the Krüppel-like TF family that was originally described with repressive transcriptional function (7). Numerous reports have demonstrated that ZNF217 can influence endogenous signaling networks governing hallmarks of cancer (8, 9) including sustained proliferation, invasion/metastasis, and resistance to chemotherapy-induced cell death in other solid tumor types such as ovarian and breast cancer (10–13). This emerging evidence suggests TFs could represent novel candidates for therapy in osteosarcoma (14). However, many of these TFs are not currently directly targetable demonstrating the necessity of identifying upstream mediators regulating their biological function.

In this report, we describe a distinct oncogenic role for ZNF217 in accelerating osteosarcoma development, tumor growth, and metastasis in part through transcriptional changes that lead to hyperactivation of PI3K–AKT signaling. Importantly, we show that blockade of AKT signaling with a clinically relevant small-molecule inhibitor [triciribine (15)] is effective at reducing tumor growth and metastasis in ZNF217⁺ tumors. Together, our results suggest continued preclinical evaluation of triciribine as a novel therapy for patients with incurable osteosarcoma.

RNA sequencing was obtained from previously published data available elsewhere (6, 16, 17) and from the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiative, phs000218, managed by the NCI. Information about TARGET can be found at http://ocg.cancer.gov/programs/target.

Calculated segment means (SM) were obtained and analyzed from 88 osteosarcoma samples available within the TARGET initiative using Genomic Suite software (Partek). Cut-off values for categorization were: diploid (SM 1.9–2.1), deletion (SM < 1.9), and amplification (SM > 2.1).

Tissue microarray (TMA) staining and quantification was performed with methods described previously (17). Briefly, two individual osteosarcoma TMAs containing 40 samples in duplicate were stained for ZNF217 and p-AKT^Ser473, respectively, and evaluated using the HALO imaging analysis platform (Indica labs). Single cells throughout each TMA section were analyzed for the presence of ZNF217 staining and cellular location (cytoplasm or nucleus were quantified separately via software). Antibodies and other reagents used are available in Supplementary Table S1.

All osteosarcoma cell lines, immortalized osteoblasts (hFOB1.19), and HEK 293Ts were obtained and maintained in accordance with ATCC recommendations. All cell lines were cultured as described previously (17). Normal human osteoblasts (NHO) were obtained from Lonza and cultured in osteoblast growth medium (#CC-3207, Lonza). With the exception of hFOB1.19 and NHOs, which were recently purchased, other cell lines were authenticated by the University of Arizona Genetics Core using short tandem repeat profiling.

ZNF217 (#M-004987-00-0005, Dharmacon) and/or control nonsilencing (#D-001206-14-05, Dharmacon) pooled siRNAs were used for all transient knockdown studies. Osteosarcoma cells were transfected at a final working concentration of 17 nmol/L using RNAiMAX (#13778150, Thermo Fisher Scientific) and all analyses were performed 48 hours posttransfection. Stable knockdown and overexpression of ZNF217 were achieved with methods described previously (17). Briefly, stable knockdown of ZNF217 was accomplished with pGIPZ lentiviral vectors expressing an short hairpin (shRNA) against ZNF217 in conjunction with GFP and a puromycin selection marker (shZNF217: #V2LHS_196547, Open Biosystems). Control pGIPZ vector (shCON) with nontargeting shRNA was used as a control (#RHS4346, Open Biosystems). Stable overexpression was achieved via a lentiviral ZNF217 vector (#98384, Addgene) or empty vector control (#41392, Addgene). Lentiviral particles were produced with HEK 293T cells cotransfected with shRNA- or overexpression-containing vectors, pMD2.G envelope (#12259, Addgene), and psPAX2 (#12260, Addgene) packaging vectors. Viruses were concentrated with Lenti-X (#631232, Clontech) and stable lines were established via puromycin selection at 1 μg/mL following viral transduction.

qRT-PCR was performed as described previously (17). Following RNA extraction, 1 μg of total RNA was reverse transcribed into cDNA (#04379012001, Roche) and qRT-PCR was performed in triplicate using SYBR green mix (#4472908, Thermo Fisher Scientific) on a CFX96 Touch System (Bio-Rad). All measurements were calculated using the ΔΔC_T method.

• ZNF217 Forward: 5′-GATGTTACTCCTCCTCCGGATG-3′

• ZNF217 Reverse: 5′-CACACTTGGCCTGTATCTGCA-3′

• ACTB Forward: 5′-CACAGGGGAGGTGATAGCAT-3′

• ACTB Reverse: 5′-CTCAAGTTGGGGGCACAAAA-3′

Western blotting was performed as described previously (17). Briefly, total lysates (∼20–30 μg) were transferred to PVDF membranes, incubated with primary and secondary antibodies, developed, and imaged using a LI-COR Odyessy. All RNAi and drug experiments were performed for 48 hours prior to Western blot analyses. A complete list of antibodies and other reagents utilized is available in Supplementary Table S1.

IHC and hematoxylin and eosin (H&E) staining procedures were performed as reported previously (6). Formalin-fixed and paraffin-embedded tissue was sectioned at 4 μmol/L. A complete list of antibodies and other reagents utilized is available in Supplementary Table S1.

Osteosarcoma cells (SJSA-1 and HOS) were plated onto sterile glass coverslips and incubated overnight at 37°C in normal media. Cells were washed with PBS and fixed with 10% neutral buffered formalin for 10 minutes at room temperature. Following 3× washing with PBS, cells were permeabilized in 0.1% Triton X-100/PBS (#T8787, Sigma-Aldrich) for 10 minutes. After 3× washes in PBS, cells were blocked for 1 hour at room temperature in 4% FBS/PBS followed by incubation in 1° antibody/blocking buffer overnight at 4°C. Post 1° incubation, cells were washed 3× in PBS and then incubated in conjugated 2° antibody/blocking buffer for 2 hours at room temperature. Following a gentle PBS wash, antifade mountant containing DAPI (#P36931, Thermo Fisher Scientific) was added, coverslips were mounted on glass slides, and imaged using the Cytation 5 cell imaging reader (BioTek). Gen5 software (BioTek) was utilized to acquire and merge immunofluorescence (IF) and DAPI channels. Incubation with 2° only controls yielded no significant IF staining.

Cell fractionation (2 × 10⁶ cells/osteosarcoma cell line) was performed according to manufacturer's instructions using the subcellular protein fractionation kit for cultured cells (#78840, Thermo Fisher Scientific). Equal amounts of total lysate (15 μg) were analyzed via Western blot analysis for both osteosarcoma cell lines.

Both triciribine (#S1117, Selleckchem) and LY294002 (#S1105, Selleckchem) were prepared according to manufacturer's instructions. Both compounds were dissolved in DMSO for all in vitro experiments.

Cell growth assays were performed as described previously (6, 17). Briefly, RNAi-modified cells (1.2 × 10³ cells/well) were seeded in 96-well plates. For drug treatment experiments, cells (2 × 10³ cells/well) were incubated for 48 hours with LY294002 or DMSO. Absorbance was measured at 490 nm and 650 nm using a SynergyMx (BioTek) plate reader at 24, 48, 72, and 96 hours postplating for RNAi experiments and at 48 hours post for LY294002 experiments.

RNAi-modified or triciribine-treated cells (2.5 × 10⁴ cells/chamber) were seeded in 500 μL of serum-free media in the upper chamber of 8 μm inserts (#353097, Corning). The lower chamber was filled with 750 μL media fortified with 10% FBS as a chemoattractant. After 24 hours, nonmigrating cells where removed with a cotton swab. Migrated cells located on the lower side of the chamber were fixed with crystal violet, air dried, and photographed to quantify migration of cells.

RNAi-modified or triciribine-treated cells (1 × 10⁴ cells/well) were seeded into a 0.35% agar solution placed on top of a 0.5% agar in 6-well plates and allowed to incubate for 1–4 weeks. The resultant colonies were fixed, divided into four quadrants, and imaged using microscopy. Colonies were quantified via ImageJ v1.52a software using a standard colony quantification macro (6).

Osteosarcoma cells (2 × 10³ cells/well) were seeded into 384-well plates using a Biomek 2000 24 hours prior to experiments. Triciribine was added in quadruplicate wells per dose in a 12-point two-fold dose-response manner using the acoustic Echo 550 liquid dispenser (Labcyte). At 48 hours post triciribine treatment, cells were incubated with alamarBlue reagent (#DAL1100, Thermo Fisher Scientific) and fluorescence was read on a CLARIOstar microplate reader (BMG LABTECH). Cell viability was calculated by fluorescence of experimental wells in percent of unexposed control wells with blank values subtracted. Dose-response curves were generated using nonlinear regression log(inhibitor) versus response-variable slope model.

Apoptosis assays were performed as described previously (17). Briefly, cells were resuspended in Annexin-V binding buffer and stained with Annexin-V and 7-AAD according to manufacturer's instructions (#BDB556547, BD Pharmingen). Cells were analyzed on an LSR II or Fortessa digital flow cytometer (BD Biosciences) at the University of Minnesota Flow Cytometry Resource. Analysis was performed using FlowJo software (FlowJo, LLC).

A combination of three transgenes was used: (i) osteoblast-specific tetracycline-controlled transactivator (tTA) and Cre-recombinase driven by the Sp7 promoter [Sp7-tTA,tetO-EGFP::Cre (abbr. Sp7-tTA/Cre)]; (ii) conditionally expressed Trp53^lslR270H/+ dominant-negative (D/N) allele; and (iii) tetracycline responsive promoter element (TRE)-driven human ZNF217 with Luciferase reporter (TRE-hZNF217-IRES-Luciferase; abbr. TRE-hZNF217/Luc). Animals expressing #1 or #2 transgenes outlined above were generated previously and described in detail elsewhere (6). A PB/SB-TRE-GOI-IRES-Luciferase–based vector (6, 17) containing a human ZNF217 cDNA (Open Biosystems) was generated via the LR cloning system (#11791020, Thermo Fisher Scientific) for #3 prior to linearization with SphI (#R0182, New England Biolabs) and pronuclear injection described in the following section.

Transgenic TRE-hZNF217-IRES-Luciferase mice were generated by University of Minnesota Mouse Genetics Laboratory. C57Bl/6J (Jackson Laboratories) females, 21–28 days of age, were superovulated for synchronized fertilized one cell embryo production. Briefly, mice were housed under a 12-hour/12-hour light/dark cycle. Females were injected with five international units (IU) of pregnant mare serum intraperitoneally at 1:00 PM on day 1. The same females were injected with 5 IUs of human chorionic gonadotropin intraperitoneally at 12:00 PM on day 3 and immediately mated 1:1 with C57Bl/6J males. Females were checked for copulation plugs on the morning of day 4. Mice with copulation plugs were sacrificed for embryo harvest. Fertilized embryos underwent pronuclear injection on a Leica DM4 microscope. Injected embryos were implanted into 0.5 days postcopulation pseudopregnant ICR females (Charles River Laboratories), 25 embryos per recipient. Resulting pups were born 19.5 days after implantation. Tail clips were genotyped following weaning at 21 days of age.

Small tail biopsies from weaned animals were collected and genomic DNA was extracted using phenol:chloroform:isoamyl alcohol extraction following overnight digestion in SDS extraction buffer (6). PCR was performed using GoTaq green master mix (#M7121, Promega). Amplicons were resolved on 1% agarose gels and analyzed for the presence of transgenes.

• TRE-hZNF217-IRES-Luciferase Forward: 5′-TCCAGCTCGACGTTAGAAGG-3′

• TRE-hZNF217-IRES-Luciferase Reverse: 5′-AGGAACTGCTTCCTTCACGA-3′

• Trp53^lslR270/+ Forward: 5′-TTACACATCCAGCCTCTGTGG-3′

• Trp53^lslR270H/+ Reverse: 5′-CTTGGAGACATAGCCACACTG-3′

• Trp53^lslR270H/+ LSL: 5′-AGCTAGCCACCATGGCTTGAGTAAGTCTGCA-3′

• Sp7-tTA,tetO-EGFP::Cre Forward: 5′-CTCTTCATGAGGAGGACCCT-3′

• Sp7-tTA,tetO-EGFP::Cre Reverse: 5′-GCCAGGCAGGTGCCTGGACAT-3′

Suspension cultures from fresh mouse spleen cells were initiated: after 48 hours incubation with Concanavalin A, cells were harvested using standard cytogenetic methods (colcemid arrest, followed by treatment with 0.75 mol/L KCl hypotonic solution, and fixation with 3:1 methanol:acetic acid). Harvested cells were spread onto glass slides. ZNF217 DNA probes were labeled by nick translation reaction (Nick Translation Kit - Abbott Molecular) using Orange 552 dUTP (Enzo Life Science). Sizes of the nick translated fragments are checked by electrophoresis on a 1% TBE gel. The labeled DNA is precipitated in COT-1 DNA, salmon sperm DNA, sodium acetate, and 95% ethanol, then dried and resuspended in 50% formamide hybridization buffer. The probe/hybridization buffer mix and slide were denatured, probe was applied to the slide, and slide was hybridized for 48 hours at 37°C in a humidified chamber. After hybridization, the FISH slides were washed in a 2×SSC solution at 72° for 15 seconds, and counterstained with DAPI stain. Fluorescent signals were visualized on an Olympus BX61 microscope workstation (Applied Spectral Imaging) with DAPI and Texas Red filter sets. FISH images were captured using an interferometer-based charge-coupled device (CCD) cooled camera (ASI) and FISHView ASI software.

Animals were injected intraperitoneally with luciferin (#XR-1001, Xenogen) in sterile PBS at 10 μL/g body weight 10–15 minutes prior to imaging. Following administration, animals were gently anesthetized using a mixture of isoflurane and oxygen and imaged using the IVIS 50 in vivo bioluminescent system. Functional Luciferase imaging was conducted during initial genetically engineered mouse model (GEMM) generation to ensure proper cassette function prior to establishing cohorts.

Upon first observation of osteosarcoma tumor, time to development was calculated (expressed as % osteosarcoma-free survival) and animals were euthanized according to Institutional Animal Care and Use Committee (IACUC)/institutional regulations. Upon euthanasia, animals were again visually inspected for tumors. Both the number of tumors and location were recorded prior to necropsy.

A total of 2 × 75 bp FastQ paired-end reads (n = 30 million average/sample) were trimmed using Trimmomatic (v0.33) enabled with the optional “-q” option; 3 bp sliding-window trimming from 3′ end requiring minimum Q30. Quality control of raw sequence data for each sample was performed with FastQC. Read mapping was performed via Hisat2 (v2.1.0) using the human genome (GRCh38) as a reference. Gene quantification was done via Feature Counts (Subread package) for raw read counts. Differentially expressed genes (DEG) were identified using the edgeR (17) feature in CLCGWB (Qiagen) using raw read counts. A heatmap was generated using 440 DEGs (fold change ± 2, FDR < 0.05). STRINGDB (https://string-db.org) was used for pathway enrichment analysis using all DEGs identified. Raw RNA-sequencing data from this study are available via Gene Expression Omnibus (GSE147413).

All animal procedures were performed in accordance with protocols approved at the University of Minnesota (Minneapolis, MN; #1905–37099A) in conjunction with the ACUC. Wildtype or lentiviral-modified osteosarcoma cells (2.5 × 10⁵ in PBS, 20 μL/injection) were implanted through intratibial injection in 6–8 weeks old male and female immunocompromised mice [NOD Rag Gamma, Jackson Labs (18, 19)], using a sterile 29-gauge, 0.3 mL insulin syringe (#8881600145, Covidien) just above the calcaneus. Tumor volume was calculated via caliper measurements using the formula V = (W*W*L)/2 where V equals tumor volume, W equals tumor width, and L equals tumor length (20). To comply with institutional regulations, all animal experiments were ended when tumor volume approached the 800–1,000 mm³ threshold (postimplantation day 31, PID31)

Following in vivo endpoint [postimplantation day 31 (PID 31)], whole lungs were subjected to H&E staining. Two slides containing 2–3 lung slices interspaced spaced throughout the whole tissue were made per animal (n = 3/group indicated). A total of 4 representative fields of view were captured per animal and the number of micrometastatic nodules as well as micrometastatic area of nodules observed were quantified and averaged (n = 24 images/group indicated).

Once orthotopic osteosarcoma tumors were established (PID 10), animals were randomized and enrolled onto study. Previous clinical research indicated that adult patients with cancer received doses of triciribine between 20–48 mg²/m²/day, however; cumulative toxicity was noted in adult patients receiving 30 mg²/m²/day [∼10 mg/kg/day in mice (21)] despite no toxicities with a 45 mg/m² single administration (22). Given the majority of osteosarcoma cases are in adolescence (1) and that higher dosages can be required for efficacy in pediatric patients (23, 24), we rationalized that a less frequent, higher-dose regimen of triciribine may overcome these clinical limitations to date in adults and better serve future pediatric patients. As such, animals received 40 mg/kg triciribine or control treatment three times weekly (3×) intraperitoneal beginning PID 10. Triciribine was prepared in a 1% DMSO/30% polyethylene glycol/1% Tween-80 solution for all in vivo studies.

All statistical analyses were performed using Prism v8 software (GraphPad). Two groups were compared using a two-tailed unpaired Student t test with or without Welch correction. Three or more groups were compared using one-way ANOVA or two-way ANOVA analyses with Bonferroni post hoc. A log-rank test was utilized for animal survival curve analysis. Simple linear regression analysis was utilized to test for significance between slopes/intercepts. All statistical analyses are individually indicated throughout in figure legends. In all cases, P < 0.05 was considered statistically significant.

Comparative genomics analysis in our previous work uncovered copy-number variation (CNV) in SB-predicted oncogene ZNF217 across human osteosarcoma samples (6). Further substantiating these findings in a second dataset (TARGET) recently made available since our original report, ZNF217 was found to be amplified in 66% of samples and this amplification positively correlated with increased expression of ZNF217 (Fig. 1A). Next, we evaluated ZNF217 expression across a series of osteosarcoma patient samples in duplicate using IHC staining. Positive ZNF217 staining was detectable in 51.2% of samples (41/80; Fig. 1B). As expected, ZNF217 was found in the nucleus of positive cells, but was also detected in the cytoplasm (Fig. 1B and C; Supplementary Fig. S1). Likewise, ZNF217 was also found localized in both the nucleus and the cytoplasm of cell lines (Supplementary Fig. S2A and S2B). Finally, to examine the expression and function of ZNF217 in osteosarcoma, we first probed four osteosarcoma cell lines for ZNF217 via Western blot analysis and compared them with NHOs. ZNF217 was detectable and aberrantly expressed in all osteosarcoma cell lines and not detected in NHOs (Fig. 1D). Overall, these data support the existence of ZNF217⁺ cells in human samples and osteosarcoma cell lines localized in both the nuclear and cytoplasmic compartments.

Given the fact that a subset of SB-mutagenized tumors were driven by ZNF217 insertions in our previous work (6) and its widespread presence in human tissue samples and cell lines (Fig. 1), we hypothesized that ZNF217 has an oncogenic role in osteosarcomagenesis. To study this hypothesis, we generated a novel GEMM whereby we achieved osteoblast-specific (Sp7-tTA,tetO-EGFP::Cre) expression of human ZNF217 (TRE-hZNF217-IRES-Luciferase) and evaluated the capacity of hZNF217 in osteosarcoma initiation and progression in the presence or absence of an osteosarcoma relevant predisposing tumor suppressor allele Trp53^lslR270H/+ (Fig. 2A). Positive TRE-hZNF217-IRES-Luciferase transgene integration was confirmed via FISH (Fig. 2B) and genotyping of all transgenes was confirmed via PCR (Fig. 2C). Functional Luciferase expression was validated in vivo using bioluminescent imaging (Fig. 2D). Human ZNF217 overexpression on the Trp53^lslR270H/+ background significantly accelerated osteosarcoma development (median 14.3 vs. 21.2 months) and penetrance (69% vs. 50% vs. 0%) as compared with Trp53^lslR270H/+;Sp7-tTA/Cre and TRE-hZNF217/Luc;Sp7-tTA/Cre animals, respectively (Fig. 2E). No osteosarcomas were observed in TRE-hZNF217/Luc;Sp7-tTA/Cre animals lacking the predisposing background (Fig. 2E, blue line). No appreciable difference in the number of osteosarcoma tumors per mouse were observed between each genotype (Fig. 2F). 3/12 control animals (25%) had visible macrometastatic nodules in the lungs [∼1–4 individual nodules/animal, in line with our previous report (6)]; however, no macroscopic metastatic nodules were observed in TRE-hZNF217/Luc;Trp53^lslR270H/+;Sp7-tTA/Cre animals. The sites of osteosarcoma development were biased to the jaw/cranium (22% versus 8%) and pelvis (22% vs. 0%) in TRE-hZNF217/Luc;Trp53^lslR270H/+;Sp7-tTA/Cre-expressing animals as compared with control animals (Fig. 2G). These findings were similar to Walkley and colleagues, where facial tumors were the most common site of mouse osteosarcoma, which typically had no evidence of metastasis (25). No osteosarcoma tumors developed in the pelvis of control animals (left pie graph) and no spinal tumors were present in TRE-hZNF217/Luc;Trp53^lslR270H/+;Sp7-tTA/Cre-expressing animals (Fig. 2G). Tumors from experimental animals were positive for human ZNF217 via Western blot analysis (Fig. 2H) and were grossly/histologically consistent with osteosarcoma via H&E staining (Fig. 2I; ref. 6). In sum, these data indicate that osteoblast specific overexpression of ZNF217 in Trp53^lslR270H/+ predisposed mice leads to accelerated osteosarcomagenesis.

As expression of human ZNF217 alone in mouse osteoblasts was not sufficient to initiate osteosarcoma development in our GEMM (Fig. 2E, blue line), we deduced that ZNF217 may be playing a larger role in osteosarcoma progression and not tumor initiation. To more closely study this, we adopted a gain/loss-of-function strategy in established osteosarcoma cell lines that better mimic facets of cellular transformation involved in tumor progression and metastasis in vitro and in vivo over our GEMM. Overexpression of ZNF217 in immortalized osteoblasts and SJSA-1 osteosarcoma cells (Fig. 3A) increased cellular proliferation (Fig. 3B). Transient pooled siRNA knockdown of ZNF217 (Fig. 3C and D) in vitro reduced cellular proliferation (Fig. 3E), cellular migration (Fig. 3F), and colony formation in soft agar (Fig. 3G).

With the profound alterations in transformed phenotypes associated with osteosarcoma malignancy in ZNF217 overexpression and knockdown cell lines, we next asked how ZNF217 expression influences the osteosarcoma transcriptome to promote those malignant phenotypes by subjecting SJSA-1 cells to RNA sequencing following transient ZNF217 knockdown and analyzing the expression profiles. We obtained a total of 440 DEGs in our dataset (Supplementary Fig. S3A; fold change ± 2, FDR < 0.05) which consisted of 242 upregulated and 198 downregulated DEGs. All DEGs identified are shown in Supplementary Table S2. A functional protein network coded by all identified DEGs was assembled using STRINGDB (https://string-db.org; Supplementary Fig. S3B). To identify biological attributes and functional categories of the DEGs identified, Gene Ontology (Biological process GO; refs. 26, 27) and Reactome (28) pathway analyses were performed on all 440 DEGs altered using STRINGDB. Among the most significantly enriched GO terms were DEGs involved in extracellular matrix organization (ECM), response to stress/stimuli, regulation of cell death, and cell migration (Supplementary Fig. S3C). Significant Reactome pathways enriched included ECM organization, collagen modification, integrin signaling, platelet-derived growth factor signaling, and steroid metabolism (Supplementary Fig. S3D).

Because metastasis is strongly linked to these phenotypes, we further asked whether ZNF217 modulates metastasis using SJSA-1 cells with the highest ZNF217 expression (Fig. 1D) in vivo following establishment of a stable shZNF217 cell line (Fig. 3H). Similar to our in vitro experiments and RNA-sequencing findings, shRNA knockdown of ZNF217 led to reduced osteosarcoma tumor growth (Fig. 3I and J) and metastasis (Fig. 3K–M). Together, these data support a role for ZNF217 in osteosarcoma tumor growth and metastasis.

Previous work from Littlepage and colleagues in breast cancer suggests ZNF217 contributes to numerous facets of transformation also important in osteosarcoma biology, including epithelial-to-mesenchymal transition, cytoskeletal rearrangement, tumor progression, and metastasis modulated in part through PI3K–AKT signaling (13, 29–31). With this is mind, we sought to investigate the PI3K–AKT signaling pathway further in the context of osteosarcoma (Fig. 4). PI3K class I catalytic enzymes have been widely implicated in cancer and represent a diverse class with biological and clinical relevance (32). Given this, we correlated ZNF217 expression with the four class I genes PIK3CA, PIK3CB, PIK3CD, and PIK3CG and found positive correlations across all genes in osteosarcoma tumor samples combined from two independent datasets (Fig. 4A). In further support, treatment with PI3K inhibitor LY294002 reduced ZNF217 expression (Fig. 4B), reduced cellular proliferation (Supplementary Fig. S4A), and induced modest apoptosis (Supplementary Fig. S4B). Similar to PI3K class I genes, ZNF217 also positively correlated with expression of all three AKT isoforms AKT1, AKT2, and AKT3 (Fig. 4C). In addition, active AKT was found to be highly abundant (96.3%) in human osteosarcoma samples (Fig. 4D) much like ZNF217 in Fig. 1C and D, overexpression or knockdown of ZNF217 increased/reduced p-AKT^Ser473 in osteosarcoma cell lines (Fig. 4E), and gene deficiency led to increased apoptosis similar to PI3K inhibition (Supplementary Fig. S5; Supplementary Fig. S4B, respectively). Together, these alterations suggest that ZNF217 promotes a highly invasive and proliferative signaling program that enhances osteosarcoma malignancy and metastasis in part through regulation of PI3K–AKT signaling.

TFs account for approximately 20% of all oncogenes identified to date and historically have been difficult to drug directly (33). Despite this challenge, recent studies have sought to identify candidate small molecules and miRNA-based therapeutics for combating ZNF217-induced oncogenic signaling, particularly in breast cancer (29). To date, one such compound, triciribine, has been identified and shown to effectively reduce tumor growth/metastases in ZNF217⁺ breast tumor cells (31) and effective at overcoming Zfp217-induced breast cancer chemotherapy resistance when in combination with paclitaxel (34). We hypothesized that triciribine treatment would be an effective therapy for combating the ZNF217–PI3K–AKT positive feedforward loop we identified in osteosarcoma cells (Fig. 4). Triciribine treatment inhibited AKT^Ser473 and total AKT levels (Fig. 5A), cell viability (Fig. 5B), cellular migration (Fig. 5C), and soft agar colony formation (Fig. 5D). Similar to our PI3K inhibitor studies (Supplementary Fig. S4B; Fig. 4B), triciribine induced apoptosis (Fig. 5E; Supplementary Fig. S6) and reduced ZNF217 expression (Fig. 5F). These data suggest that targeting the PI3K–AKT signaling pathway has robust cytotoxic effects on ZNF217-expressing osteosarcoma cell lines treated in vitro and that ZNF217 is part of a feedforward loop involving PI3K–AKT signaling.

Following our in vitro drug studies, we sought to examine the therapeutic potential of triciribine in an orthotopic mouse model of osteosarcoma using two different osteosarcoma cell lines. Tumor-bearing animals were randomized and treated three times weekly (3×) with triciribine or control beginning PID 10. Tumor growth was measured until endpoint (PID 31). Tumor volumes were significantly reduced in animals receiving triciribine (Fig. 6A). Triciribine treatment resulted in 5.4- and 3.0-fold changes in average tumor volume at endpoint in animals as compared with control counterparts (SJSA-1 and HOS-injected animals, respectively). Similarly, triciribine treatment resulted in 2.3- and 2.9-fold reductions in tumor weight at endpoint (Fig. 6B, SJSA-1 and HOS, respectively). Pharmacodynamic analysis of tumor tissue confirmed that cytoplasmic ZNF217 expression (and to a lesser extent nuclear expression) was reduced in triciribine -treated animals (Supplementary Fig. S7) similar to our Western blot analysis in vitro findings (Fig. 5F). Triciribine treatment was also associated with reduced micrometastatic nodule formation and area (Fig. 6C–E). In sum, AKT signaling blockade through triciribine treatment effectively reduces ZNF217⁺ osteosarcoma tumor growth and metastasis (Fig. 7).

It is undeniable that novel treatment options are urgently needed for patients with metastatic osteosarcoma. Here we find that ZNF217 is responsible for the regulation of a feedforward mechanism involving PI3K–AKT signaling that ultimately represents a therapeutically relevant target for significantly reducing osteosarcoma tumor growth and metastasis.

We recently identified ZNF217 (murine Zfp217) as a putative driver for a subset of SB-induced osteosarcomas (6). Our new autochthonous model of osteosarcoma indicates a role for ZNF217 in the progression, not initiation, of osteosarcoma. Encouragingly, other ZNF family members have also been implicated in osteoblastic differentiation (35), facilitating commitment to the osteoblast lineage (36) and were found to be putative drivers in some of our SB-induced osteosarcomas (ref. 6; ZNF592). ZNF217 was amplified in 66% of samples examined leading to ectopic oncogenic expression at both the RNA and protein level in osteosarcoma tumor samples; observations in line with our inability to detect it in NHOs. Both primary human osteosarcoma samples and cell lines had detectable ZNF217 protein in nuclear and cytoplasmic compartments and ZNF217 was found to be amplified in a significant percentage of osteosarcoma samples as previously reported in breast cancer (29, 31). Interestingly, cytoplasmic ZNF217 expression has recently been reported in solid tumors of the breast and colon (37, 38) and has been functionally linked to estrogen receptor alpha trafficking in MCF-7 breast cancer cells (39, 40). In combination with our findings, further investigation into the mechanisms of ZNF217 localization as they relate to osteosarcoma tumor development, progression, and metastasis are needed.

Given the role of ZNF217 in osteosarcoma progression found in our GEMM, it is not surprising that genetic knockdown of ZNF217 significantly inhibited hallmarks of cellular transformation, ultimately culminating in reduced osteosarcoma tumor growth and metastasis in our orthotopic mouse model. Recent reports suggest that ZNF217 can enhance metastatic bone growth in breast cancer and promote the formation of osteolytic lesions in part through regulation of bone morphogenic protein (BMP) signaling (41, 42). BMP signaling has been well documented in normal bone processes of growth and differentiation (43, 44). BMPs are abundant in osteosarcoma and have been shown to correlate with metastasis (45, 46). This is further supported by complementary reports of oncogenic ZNF217 signaling in colon (47), prostate (48), breast (31), and ovarian (11) solid tumors. Moreover, ZNF217 also protected cells from death, as knockdown readily induced apoptosis. This suggests ZNF217 promotes progression and metastasis in part through regulation of PI3K–AKT survival signaling. Our group and others have demonstrated dysregulated PI3K–AKT signaling can drive a subset of murine (6), canine (49, 50), and human osteosarcomas (51–53). This dysregulation is present in a majority of localized osteosarcoma cases and in all advanced-stage cases (54, 55). Dual reports from Milne and colleagues and Feng and colleagues elegantly showed that AKT signaling plays a critical role in murine double minute 2 (MDM2) phosphorylation resulting in protein stabilization and disruption of TP53 signaling (56, 57). Mirroring these previous studies, we found correlation between ZNF217 expression and catalytic PI3K genes as well as downstream AKT1, AKT2, and AKT3 transcripts in osteosarcoma patient RNA-sequencing data. Consistent with these findings, we found abundant active AKT staining (p-AKT^Ser473) in our osteosarcoma TMA. MDM2 is known to interact with ZNF217 (58) and is amplified in metastatic osteosarcoma (59) suggesting a role for ZNF217-induced PI3K–AKT signaling. Given our identification of PI3K–AKT signaling as a therapeutically targetable axis in ZNF217-expressing osteosarcoma cells, these data provide an attractive rationale for investigating small molecule AKT inhibitors preclinically for the treatment of ZNF217⁺ osteosarcomas.

In support of our genetic findings, cytotoxic triciribine treatment significantly reduced p-AKT^Ser473, phenotypes of cellular transformation in vitro, ZNF217 expression, and tumor growth/metastasis in vivo. We also found that treatment of osteosarcoma cells with PI3K inhibitor LY294002 inhibited cell viability, produced modest apoptosis, and reduced ZNF217 protein expression [unlike a previous report (31)]. One possible reason for this discrepancy could be that here we show triciribine treatment reduces basal ZNF217 expression, while Littlepage and colleagues examined ZNF217 expression following cell starvation, treatment with triciribine, and in the presence of heregulin/neuregulin-1β [see Fig. 7D (31)]. In sum, these data suggest a positive feedback mechanism for ZNF217-induced AKT signaling which agrees with our overexpression studies that demonstrate strong activation of AKT (p-AKT^Ser473) in ZNF217-overexpressing immortalized osteoblasts and SJSA-1 osteosarcomacells.

Both phase I and II clinical trials of triciribine have been conducted in advanced malignancies previously (22, 60–63). While treatment was associated with some side effects, including hepatotoxicity and hyperglycemia, modest benefits in stabilizing disease were observed in at least one patient with breast cancer undergoing treatment (62). A new form of triciribine (TCN-P) has been developed (PTX-200, Prescient Therapeutics) and is currently recruiting patients with refractory or relapsed acute leukemia (NCT02930109) and has active status for phase I treatment of ovarian (NCT01690468) and phase I/II treatment of stages II–IV breast cancer (NCT01697293) in combination with chemotherapy.

In summary, our data demonstrate a role for ZNF217 in promoting osteosarcoma development and orthotopic growth/metastasis through modulation of an oncogenic gene network regulated in part through active PI3K–AKT signaling. Importantly, our work has uncovered the therapeutic potential of targeting the PI3K–AKT signaling axis in osteosarcoma through the use of the small-molecule AKT inhibitor triciribine. Together, this supports the use of triciribine in patients with osteosarcoma with ZNF217⁺ and active PI3K–AKT tumors.

G.M. Draper reports grants from NIH during the conduct of the study. D.A. Largaespada reports grants from the American Cancer Society during the conduct of the study; grants from Genentech; personal fees and other from BmoGen Biotechnologies, Inc. (salary and equity); and other from NeoClone Biotechnologies, Inc. (equity), Recombinetics, Inc. (equity options), Luminary Therapeutics, Inc. (equity), and ImmuSoft Inc. (equity) outside the submitted work. No disclosures were reported by the other authors.

B.A. Smeester: Conceptualization, data curation, formal analysis, methodology, writing-original draft, writing-review and editing. G.M. Draper: Conceptualization, data curation, formal analysis, methodology, writing-original draft, writing-review and editing. N.J. Slipek: Data curation, writing-original draft, writing-review and editing. A.T. Larsson: Data curation, writing-original draft, writing-review and editing. N. Stratton: Data curation, writing-original draft, writing-review and editing. E.J. Pomeroy: Data curation, writing-original draft, writing-review and editing. K.L. Becklin: Data curation, writing-original draft, writing-review and editing. K. Yamamoto: Data curation, writing-original draft, writing-review and editing. K.B. Williams: Data curation, writing-original draft, writing-review and editing. K. Laoharawee: Data curation, writing-original draft, writing-review and editing. J.J. Peterson: Data curation, writing-original draft, writing-review and editing. J.E. Abrahante: Data curation, writing-original draft, writing-review and editing. S.K. Rathe: Data curation, writing-original draft, writing-review and editing. L.J. Mills: Data curation, writing-original draft, writing-review and editing. M.R. Crosby: Data curation, writing-original draft, writing-review and editing. W.A. Hudson: Data curation, writing-original draft, writing-review and editing. E.P. Rahrmann: Data curation, writing-original draft, writing-review and editing. D.A. Largaespada: Conceptualization, resources, supervision, funding acquisition, writing-original draft, writing-review and editing. B.S. Moriarity: Conceptualization, resources, supervision, funding acquisition, writing-original draft, writing-review and editing.

The authors would like to thank the Clinical and Translational Science Institute Histology and Research Laboratory team member Colleen Forester for tissue preparation and histology services. The authors acknowledge the Minnesota Supercomputing Institute, the Institute for Therapeutics Discovery and Development, and the Mouse Genetics Laboratory at the University of Minnesota for providing resources that contributed to the research results reported within this article. The cytogenetic analyses were performed in the Cytogenomics Shared Resource at the University of Minnesota with support from the comprehensive Masonic Cancer Center NIH Grant no. P30 CA077598. B.A. Smeester. was previously supported by an NIH NIAMS T32 AR050938 Musculoskeletal Training Grant and is currently supported by a Doctoral Dissertation Fellowship through the Graduate School at the University of Minnesota. G.M. Draper is supported by an NIH NIGMS T32 GM113846-09 Stem Cell Biology Training Grant. E.J. Pomeroy is supported by an NIH NIAID T32 AI997313 Immunology Training Grant. K.B. Williams is supported by a Children's Tumor Foundation Young Investigator Award from the NF Research Initiative at Boston Children's Hospital, made possible by an anonymous gift. This work was made possible through funding from the Zach Sobiech Osteosarcoma Fund Award, Randy Shaver Cancer and Community Fund, Aflac-AACR Career Development Award, and the Children's Cancer Research Fund to B.S. Moriarity and the American Cancer Society Professor award to D.A. Largaespada.

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.