Removing IGF2 lowers but does not stop bone cancer growth in lab dishes and living models

Human osteosarcoma Saos2 cells were obtained from the American Type Culture Collection (ATCC, HTB‐85) and were maintained in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37 °C in a humidified incubator with 5% CO₂.

CRISPR‐Cas9 was used to generate IGF2 knockout (KO) cells. sgRNA off‐target effects were predicted using CHOPCHOP [34]. Two single guide RNAs targeting IGF2 at two positions on chromosome 11 (CGCGGTCCCCACAGACGAAC, at 2135379‐2135398; and CAGGTCATAATGCCGAGGTG at 2135686‐2135705) (Fig. 1) were delivered as ribonucleoprotein complexes via nucleofection using the Lonza 4D‐Nucleofector system. Transfections were performed following the manufacturer's recommended program for Saos‐2 cells (program DS‐150) and using the provided nucleofection solution. A pooled KO population was initially established, followed by single‐cell cloning to isolate individual knockout clones. To verify the knockout cell lines, genomic DNA was amplified via PCR using two distinct primer sets. The first set comprised an external primer pair flanking the deletion region (forward: 5′‐GATCATTCACCGAACGCACG‐3′, reverse: 5′‐ATGCACATGCTCTGTAGGGG‐3′), yielding a ~1 kb amplicon for wild‐type (WT) cells and a 739 bp amplicon for knockout (KO) cells. The second set consisted of an internal primer pair (forward: 5′‐AGAAGCACCAGCATCGACTT‐3′, reverse: 5′‐TGAGAAGTGGCGATGTGACG‐3′), with one primer situated within the deleted sequence to generate a 457 bp amplicon exclusively from WT cells (Fig. 1A). PCR products were subsequently analyzed by agarose gel electrophoresis: 2 μL aliquots of each PCR mixture were blended with 6× loading dye and resolved on a 1.5% (w/v) agarose gel prepared in 1× TAE buffer containing SYBR Safe nucleic acid stain. Electrophoresis was conducted at 90 V for 60 min, and a 100 bp DNA ladder was loaded concurrently as a molecular weight marker. DNA fragments were visualized under blue‐light illumination and documented using an Azure 200 gel imaging system (Fig. 1B). Successful deletion of IGF2 was further confirmed by Sanger sequencing (Fig. 1C).

Wild‐type (WT) and IGF2 knockout (KO) cells were seeded at equal densities in triplicate and grown under identical conditions. To test the effect of exogenous stimulation, various FBS concentrations or recombinant human IGF2 (R&D Systems #292‐G2) were added. Cells were harvested simultaneously when at least one population became confluent (approximately 5 days). Cell suspensions were mixed 1:1 with 0.4% trypan blue, and viable (trypan blue–excluding) cells were counted manually using a hemocytometer under brightfield microscopy. For each sample, 4 of the grid squares were counted, and the average was used to calculate cell density. Counts were performed by one blinded observer in triplicate biological replicates.

NOD.Cg‐Prkdc^scid Il2rg^tm1Wjl^/SzJ (NSG) mice (6–8 weeks old, female) obtained from The Jackson Laboratory (stock #005557) and maintained under pathogen‐free conditions in ventilated cages with nesting material in a light/dark cycle at 22 ± 2 °C and 40–60% humidity, with ad libitum access to standard irradiated rodent chow and filtered water. Mice were acclimatized for 1 week and allocated to WT or IGF2 KO groups (n = 6 per group) at random. WT or IGF2 KO Saos‐2 cells (5 × 10⁶ cells in 100 μL PBS) were injected subcutaneously into the right flank. Tumor volume, the primary outcome measure, was calculated as (length × width²)/2 from caliper measurements taken every 3–4 days. All animal procedures were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) of the Pennsylvania State University; the protocol was approved under permit number PROTO201901141. Humane endpoints were predefined and applied to all animals. Euthanasia was performed promptly when any of the following criteria were met: tumor size reaching an average diameter of 1.5 cm or a tumor volume exceeding 10% of body weight; body weight loss ≥20% compared to pretreatment baseline; body condition score <2; labored respiration; tumors interfering with normal locomotion; inability to access food or water; signs of dehydration; tumor ulceration with serous or bloody exudate; or tumor necrosis with exposure of internal tumor tissue. Once animals reached endpoint criteria, euthanasia was performed within the same day of assessment by gradual‐fill CO₂ exposure (30–70% chamber volume displaced per minute) using a compressed gas cylinder with a calibrated flow regulator; death was confirmed by cessation of respiration and heartbeat, followed by cervical dislocation performed by trained personnel. No animals died before meeting criteria for euthanasia.

Proteins annotated as growth factors (GOA: ‘growth factor activity [8083]’) and secreted mitogenic factors (GOA: ‘positive regulation of cell population proliferation [8284]’ or GOA: ‘negative regulation of apoptotic process [43066]’) with subcellular localization ‘Secreted’ (SL‐0243), excluding membrane proteins (SL‐0162) were identified in UniProt. Ligand–receptor pairs were obtained from the CellPhoneDB database (https://www.cellphonedb.org↗) [35]. Only simple one‐to‐one ligand–receptor pairs were considered; complex multimeric interactions were excluded.

Gene expression for candidate growth factors was obtained from the CCLE dataset in DepMap [31, 32] (release 23Q4) and reported as log₂(TPM + 1). Gene dependency was quantified using DepMap probability of dependency values, where higher scores indicate a higher likelihood of essentiality in a given cell line.

One WT and one IGF2‐KO Saos2 sample were submitted to Plasmidsaurus (Eugene, OR, USA) for 3′‐end RNA sequencing on an Illumina platform (single‐end, strand‐specific reads). Reads were filtered and trimmed using FastP v0.24.0, aligned to the human reference genome using STAR v2.7, and deduplicated using unique molecular identifiers (UMIs) with UMICollapse v1.1.0. Gene‐level counts were generated using featureCounts (subread v2.1.1) and normalized using the trimmed mean of M‐values (TMM) method. Functional enrichment was performed using single‐sample Gene Set Enrichment Analysis (ssGSEA) [36, 37] implemented via GSEApy v0.12 [38], against the MSigDB Hallmark gene set collection [39, 40]. As RNA sequencing was performed on a single sample per genotype, ssGSEA enrichment scores are reported descriptively; conventional phenotype‐permutation NES/FDR statistics, which require multiple replicates per group, could not be calculated. Quality control metrics, fold change values, and enrichment results are provided in the Data S1–S5.

To investigate the role of IGF2 in Saos2 cells, we designed single guide RNAs (sgRNAs) for CRISPR‐Cas9‐mediated knockout. Following single‐clone selection, we obtained three homozygous IGF2 knockout (KO) cell lines carrying a ∼306 bp deletion including the first common exon of all IGF2 isoforms annotated in the Cancer Cell Line Encyclopedia (CCLE) [41, 42]. In all annotated isoforms except one, which is not expressed in Saos2 cells according to CCLE data, this exon is also the first translated exon. Correct genomic editing was confirmed by PCR and Sanger sequencing (Fig. 1). The two sgRNAs have no predicted off‐targets with fewer than 3 mismatches and two potential off‐targets with three mismatches at chr4:75892340 and chr6:148379242, in introns of PPEF2 and SASH1, respectively. Neither gene showed evidence of dependency in Saos‐2 cells according to DepMap Chronos scores (−0.09 and −0.04, respectively).

In vitro, IGF2 KO cells exhibited reduced proliferation compared with wild‐type (WT) Saos2 cells across all fetal bovine serum (FBS) concentrations tested, although measurable growth persisted. Both WT and KO cells demonstrated enhanced proliferation in response to increasing FBS, indicating that serum factors generally promote growth; however, increased FBS did not compensate for the loss of IGF2 (Fig. 2A). In contrast, addition of exogenous IGF2 to the culture medium significantly enhanced proliferation of KO cells and substantially restored their growth to levels comparable to WT cells (Fig. 2B–D), demonstrating that IGF2 itself is sufficient to rescue the proliferation defect and that its signaling pathway remains intact.

In vivo, IGF2 KO xenografts grew substantially slower than WT tumors (Fig. 2E). At the experimental endpoint, KO tumors were smaller in volume, yet tumor formation occurred in all animals. After 70 days, KO tumors reached only approximately 25% of the size of WT tumors, confirming that IGF2 contributes to tumor growth but is not strictly essential.

To characterize the constitutive survival landscape of Saos2 cells, we compared gene expression data with DepMap dependency scores (Fig. 3A, Table 1) for known ligand–receptor pairs (Fig. 3B). While IGF1R showed a high dependency (0.71), IGF2 showed low expression (0.80), suggesting a potentially context‐dependent survival pathway or that minimal ligand input may be sufficient to sustain signaling.

Analysis of other growth factor – receptor pairs suggests that several factors could partially substitute for IGF2 function in Saos2 cells, including ITGAV (integrin αV), LIFR (Leukemia Inhibitory Factor Receptor), IL6ST (Interleukin 6 Signal Transducer), and PDGFRA (Platelet‐Derived Growth Factor Receptor Alpha). ITGAV emerged as the most essential receptor (dependency score: 0.96), and multiple ligands capable of engaging ITGAV were highly expressed: PLAUR (expression: 8.89), TNC (8.70), FN1 (8.52), MMP2 (8.00), THBS1 (7.43), TGFB1 (7.39), THY1 (7.17), LAMC1 (6.80), FGF1 (2.56), FGF2 (1.60), and SPP1 (0.72). This suggests the presence of a robust autocrine and/or paracrine extracellular matrix (ECM) signaling network and that ITGAV‐mediated adhesion and ECM interactions are critical for cell survival and proliferation that may buffer the impact of IGF2 loss.

Analysis of cytokine signaling revealed that the LIF–LIFR and CLCF1–LIFR/IL6ST axes are moderately essential, with dependency scores of 0.63 for LIFR and 0.58 for IL6ST. LIF (expression 5.70) and CLCF1 (expression 4.14) are expressed at moderate levels, suggesting the existence of redundant or complementary autocrine loops supporting JAK/STAT‐mediated survival and proliferation. Similarly, PDGF signaling via PDGFC (expression 6.03) and PDGFB (expression 1.24) acting on PDGFRA (dependency 0.48) appears moderately essential, likely contributing to cell proliferation and motility.

Receptor tyrosine kinases, including ALK (engaged by MDK 8.05 and PTN 6.38; dependency 0.29) and AXL (GAS6 5.47; dependency 0.16), showed lower dependency, indicating supportive but noncritical roles. TGFβ and BMP pathways, involving TGFB1–3 and BMP4/BMPR1A signaling through TGFBR1 (0.25) and BMPR2 (0.21), were also lowly essential, suggesting potential modulatory functions rather than critical roles in basal survival.

Overall integration of ligand expression and receptor dependency predicts several autocrine loops: a dominant essential pathway driven by ITGAV–ECM ligands; LIF/CLCF1 binding to LIFR/IL6ST, likely supporting JAK/STAT survival signaling; PDGFC/PDGFB binding to PDGFRA, likely contributing to proliferation and motility. IGF2 binding to IGF1R. Other ligands, including GAS6–AXL and BMPs, likely play secondary, supportive roles under basal conditions. These results indicate that the cell line relies, at least in vitro, predominantly on ITGAV‐mediated integrin signaling for survival and proliferation, with moderately essential cytokine and PDGF pathways providing supportive growth signals. RTK and TGFβ/BMP pathways appear nonessential under basal conditions but may contribute under stress or in vivo.

Finally, gene set enrichment analysis (GSEA) was performed using a ranked list of all genes ordered by log₂ fold change between WT and IGF2‐knockout cells. Enrichment was evaluated across the full transcriptome without imposing fold change thresholds. This ranked‐list framework assesses whether members of predefined gene sets accumulate preferentially at the extremes of the ranked distribution, rather than relying on arbitrarily thresholded gene lists [36, 37]. Negative normalized enrichment scores (NES) indicate relative depletion of a gene set in KO cells, whereas positive NES indicate relative enrichment. Significance estimates were derived from the GSEA permutation framework and are reported as FDR‐adjusted P‐values.

Of the Hallmark gene sets evaluated, we focus on those showing statistical significance (Data S2) and relevance to the phenotypes described above. Ranked GSEA (Fig. 4) revealed coordinated depletion of angiogenesis‐related gene sets in IGF2 KO cells (NES = −1.579; FDR = 0.041), suggesting reduced representation of angiogenic programs across the transcriptome. The MYC Targets V1 and V2 gene sets showed similar negative enrichment trends, although these did not reach statistical significance. No enrichment changes were observed in gene sets directly associated with canonical IGF‐axis signaling. Likewise, no enrichment of ITGAV‐related gene sets was detected, indicating that the ITGAV‐mediated interactions described above likely represent a pre‐existing survival pathway rather than a transcriptional consequence of IGF2 loss.

Collectively, these results indicate that, while IGF2 contributes to the proliferation of Saos2 cells in vitro, and significantly enhances the growth of Saos2 xenografts, its loss can be tolerated, and compensatory growth does not appear to be mediated by major changes in canonical IGF2 pathway genes but may be supported by constitutively active ITGAV–ECM signaling, as suggested by the integration of expression and dependency data.