Morpholino-driven transcriptional blockade of Dkk-1 in experimental osteosarcoma inhibits bone damage and tumor expansion by multiple mechanisms.


 Osteosarcoma (OS) is the most common primary bone malignancy. Chemotherapy plays an essential role in OS treatment, potentially doubling 5-year event-free survival if tumor necrosis can be stimulated, but long-term treatment results in detriment to health and quality of life. The canonical Wnt inhibitor Dickkopf-1 (Dkk-1) enhances OS survival in part through upregulation of aldehyde-dehydrogenase-1A1 (ALDH1A1) which neutralizes reactive oxygen species from nutritional stress and chemotherapeutic challenge. Dkk-1 also inhibits bone repair, exacerbating formation of osteolytic lesions caused by tumor infiltration. Therefore, targeting the expression of Dkk-1 in OS could reduce tumor burden and increase susceptibility to chemotherapeutics while restoring bone repair. Herein, we report that inhibiting Dkk-1 transcription by means of a vivo morpholino (DkkMo) reduced the expansion of experimental OS tumors, preserved bone volume and architecture, and stimulated tumor necrosis. This was observed in the presence or absence of doxorubicin (DRB), and as a single agent, inhibition of tumor expansion by DkkMo was equivalent to that achieved by DRB. DkkMo stimulated apoptotic and necrotic mechanisms in tumors and appeared to deplete the tumor stroma. These results indicate that administration of DkkMo with or without chemotherapeutics can substantially improve OS outcome with respect to tumor expansion and osteolytic corruption of bone.


Introduction.
Osteosarcoma (OS) is the most common primary bone malignancy. Particularly common in pediatric patients, OS and other primary bone malignancies account for approximately 9% of pediatric cancer deaths 1 . The current standard of care for treating osteosarcoma is surgery with neoadjuvant and adjuvant chemotherapy 2 . Chemotherapeutic strategies for treating osteosarcoma generally include high-dose methotrexate, doxorubicin (DRB) and cisplatin 2 .
These agents undeniably play a positive role in OS treatment, with a reported increase in 5-year event-free survival from approximately 20-40% (surgery alone) to 50-90% with successful chemotherapeutic intervention 3 . These survival rates can be stratified into responders with greater than 90% tumor necrosis after neoadjuvant therapy versus the remainder, who have 5-year event-free survival rates of 90% and 50-60% respectively 3 . In spite of the clear benefits, long-term use of chemotherapeutic agents results in side effects which can be catastrophic to patients' overall health and quality of life with studies indicating that patients with severe side effects are more likely to be non-adherent to treatment protocols [4][5][6] . One study indicated that as many as 16% of patients discontinue adjuvant chemotherapy treatment after persistent side effects 5 . Many patients experience catastrophic pain and immobility as a result of osteolytic bone lesions (OLs), which occur due to excessive bone resorption caused by bone malignancies 7 . OLs not only cause significant pain, but also increase risk of fracture and contribute to the vicious cycle between cancer cells, osteoblasts and osteoclasts which provides the ideal environment for tumor cell propagation 8 . Reducing the dependence on chemotherapy and the OL burden would significantly improve the impact of OS treatment strategies.
Several studies report high expression of Dickkopf-Wnt-signaling-pathway-inhibitor-1 (Dkk-1) in various types of cancer, including squamous cell carcinoma, pancreatic cancer, bladder cancer, hepatoblastomas, prostate, breast, multiple myeloma (MM) and OS [9][10][11][12][13][14] . Our previous study found that constitutively high expression of human Dkk-1 in the OS cell line MOSJ-Dkk1 increases tumor growth rate and bone destruction in mouse models when compared to the parental and control lines which manifested tumors primarily as non-osteolytic osteochondral nodules 15 . The observed increase in proliferation and tumorigenicity was found to be due in part to a stress-response modulated by enhanced expression of aldehydedehydrogenase-1A1 (ALDH1A1). This response occurred through constitutive inhibition of canonical Wnt signaling (cWnt) by Dkk-1, driving the balance of Wnt signaling in favor of a non-canonical Wnt pathway (ncWnt) which upregulated ALDH1A1 expression ultimately through activation of Jun kinase (JNK). ALDH1A1 is a known member of the cellular stress response arsenal, neutralizing free radicals from metabolic stressors and chemotherapeutics 15 .
The extended family of aldehyde dehydrogenases has also been implicated as a major driver of chemoresistance and survival in cancer stem cells 16 . High levels of Dkk-1 also inhibit cWntmediated differentiation of osteoprogenitors to osteoblasts thereby preventing the repair of OLs in a range of malignant bone diseases (MBD) including OS 12,15 . Therefore, targeting the expression of Dkk-1 in OS tumors could reduce the rate of expansion and survival of tumor cells, increase susceptibility to chemotherapeutics, and restore the capacity of bone to repair itself.
Herein, we demonstrate that through inhibiting Dkk-1 transcription by means of a vivo morpholino (DkkMo), it is possible to (i) reduce the expansion of MOSJ-Dkk1 tumors in vitro and vivo, (ii) preserve bone volume and architecture in vivo, and (iii) stimulate necrosis of the tumor. DkkMo had the capacity to perform these functions in the presence or absence of DRB, and the level of tumor growth inhibition by DkkMo when administered as a single agent was equivalent to that observed by DRB at high dose. In contrast with DRB, the DkkMo morpholino did not cause weight loss in experimental animals. RNA sequencing studies indicated that 5 DkkMo stimulated cell death and necrotic mechanisms in tumors, and it also appeared to deplete elements of the tumor stroma.
Collectively, these results indicate that administration of DkkMo in the presence or absence of chemotherapeutic agents has the capacity to substantially improve outcome with respect to OS tumor expansion and osteolytic corruption of bone.  18 and OS 19 , but systemic Dkk-1 levels can be substantial 10,12 , requiring large doses of antibody. We therefore hypothesized that blockade of Dkk-1 transcription may represent a more efficient approach in these systems.

DkkMo treatment reduces aldehyde dehydrogenase 1A1 production and disrupts stress response and anti-apoptotic pathways in vitro:
In MOSJ-Dkk1 OS cells, Dkk-1 triggers a stress response pathway by inhibition of cWnt signaling, thereby activating a ncWnt pathway that facilitates JNK/Jun mediated expression of ALDH1A1 15 . It is known that ALDH1A1 is one of the major aldehyde dehydrogenases responsible for neutralization of reactive oxygen species that occur when cells are subjected to nutritional and chemotherapeutic stress [20][21][22][23] , and ALDH1A1 enhances a variety of tumorigenic characteristics in bone cancer cells such as chemoresistance, metastasis and the maintenance of a tumor initiator phenotype 20,[24][25][26][27] . It was therefore hypothesized that inhibition of Dkk-1 through the action of DkkMo may increase susceptibility to nutritional and chemotherapeutic stressors by reducing expression of ALDH1A1. To test the effect of DkkMo on ALDH1A1 expression, MOSJ-Dkk1 cells were exposed to 5 µM DkkMo for 4, 6 and 9 days and ALDH1A1 transcriptional activity was assayed by quantitative RT-PCR To gain broader insight into the potential effects of Dkk1 blockade and reduction in ALDH1A1 activity, MOSJ-Dkk1 cells at logarithmic and confluent phases of growth were exposed to DkkMo for 6 days and mRNA was recovered for high throughput sequencing (HTS).
The rationale for these culture conditions was to induce cellular stress through rapid mitosis or nutritional stress respectively. Differentially expressed (DE) sequences (by >1.5-fold) between DkkMo treated and DkkScr treated cultures were identified (Supplemental Table 1) and categorized based on gene ontology term enrichment analysis (GOTEA). In both culture conditions, the greatest degree of DE mRNA enrichment occurred in GO-term lists related to stress response, programmed cell death, and response to chemical stimuli (Fig2B&C). In the case of rapidly dividing cells, GO-terms related to chemotaxis, osteogenic differentiation and immune-chemokines were also represented (Fig2B). Co-expressed, functionally related gene modules were also calculated from the dataset and hub genes with the high highest degree of connectivity within each module were also identified (FigS2). Hub genes were plotted on volcano plots to visualize potential overlap between lists of DE genes (Fig2D&E). In response to DkkMo, ALDH1A1 was downregulated in both culture conditions, and was identified as a key hub gene, further supporting the close relationship between Dkk-1 and ALDH1A1. Hsd17b4, encoding 17β-hydroxysteroid dehydrogenase/D-3-hydroxy acyl-CoA dehydrogenase also met these criteria 28 . In confluent MOSJ-Dkk1 cells only, DkkMo caused downregulation of transcript S1pr2, encoding sphingosine-1-phosphate receptor 2 29 . Ingenuity Pathway Analysis

Dkk-1 Expression Desensitizes MOSJ cells to Doxorubicin:
DRB is commonly used alone or in combination for adjuvant treatment of OS 3 , and chemoresistance to DRB in OS has been attributed in part to elevated ALDH activity 20,30 . This observation is not surprising, given that one of the key mechanisms of DRB is the generation of mitochondrial reactive oxygen species (ROS) which destroys tumor cells but also contributes to a dangerous level of cardiotoxicity 31 .
We hypothesized that Dkk-1 expression and associated ALDH1A1 levels could increase resistance to DRB, and the blockade of Dkk-1 by DkkMo could reverse this phenomenon.  (Fig4A, B). In each case, the slopes (representing the tumor growth rate) were compared using a mixed-model for repeated measures (MMRM) approach (Fig4B, FigS1B). Fluorescence measurements indicated that combination therapy and DkkMo alone reduced the rate of tumor expansion as compared to the untreated control group but co-administration of DRB did not appear to improve the effect of DkkMo alone. DRB treatment alone reduced the average rate of tumor expansion to a degree equivalent to what was observed for DkkMo alone. The lack of observed synergy or additive contribution between DRB and DkkMo could result from DRB stimulating ALDH1A1 through an alternate pathway, but DRB alone did not increase ALDH1A1 transcription even though endogenous murine Dkk-1 was slightly upregulated (FigS1C-E). It is also noteworthy to add that DkkMo administration did not affect the weight of the mice whereas DRB toxicity resulted in significant weight loss (Fig4C).
After 2 weeks of treatment, hindlimbs harboring tumors were dissected from euthanized mice and stained with Lugol's iodine contrast agent. On detailed inspection of the axial reconstructed scans, radio-dense patches were observed in the tumor masses that were more prevalent in tumors that received DkkMo and combination treatment (Fig4D, arrowed). A convolutional neural network algorithm (CNNA) was trained to segment the tumor and the radio-dense patches on the axial reconstructions with 81.87% accuracy (97% on the validation dataset) (Fig 4D,   FigS3). When the CNNA was employed to measure the volume of radio-dense structures in all tumor specimens and normalize this to total tumor volume, there was a significant increase in tumors that were treated with DkkMo or combination (Fig4E, FigS3) Collectively, the data indicated that the radio-opaque structured were collagenized necrotic foci, and that DkkMo not only slowed tumor expansion, but also had the capacity to induce tumor necrosis in vivo.

DkkMo has the capacity to reduce bone destruction in vivo:
Several studies have demonstrated that Dkk-1 has the capacity to prevent bone repair in osteolytic lesions, thereby accelerating bone destruction and facilitating tumor expansion 10,12 . Parental MOSJ-cells generate highly 11 differentiated osteochondral tumors with limited bone involvement, but MOSJ-Dkk1 cells generate primitive, aggressive and highly osteolytic tumors in mice 15 . To examine whether DkkMo treatment could reduce the osteolytic effects of MOSJ-Dkk1 cells, bones in the tumorbearing hindlimbs of mice were scanned by µCT and volumetrically analyzed. Qualitative inspection of tibias and fibulas indicated that MOSJ-Dkk1 tumors caused significant destruction.
To explore this more closely, computational superimposition of scanned images of malformed bones onto unaffected contralateral control scans highlighted where, and to what extent, the surface topology of the bone had deviated from the wild-type form. By measuring these deviations for each voxel, a profile could be generated illustrating the frequency of voxels that had deviated outwards or inwards from the plane of the healthy bone surface. A healthy bone measurement generates a narrow profile, indicating that few voxels deviated from the plane of the contralateral control bone, and if so, by a small degree (Fig 5A, above). A deformed bone would generate a broader distribution, indicating a surface topology consisting of many voxels above and below the plane of the surface of the contralateral control bone (Fig 5A, below). When applied to tibial specimens, healthy bones generated the expected narrow profile whereas

Transcriptomic sequencing of MOSJ-Dkk1 tumors indicates that DkkMo modulates proliferative, survival and immunological processes in vivo:
To gain insight into the antitumorigenic mechanism of action of DkkMo on MOSJ-Dkk1 tumors in vivo, fixed tumors were excised and subjected to HTS. Comparison between untreated (NT) and DkkMo specimens (n=4) and comparison between DRB treated and combination specimens (n=4) were performed.
After processing, lists of transcripts with >1.5-fold differential expression were generated (Supplemental Table 2

Discussion.
The classical role of Dkk-1 is to target the LRP6/5 receptor to the Kremen2 receptor for internalization and destruction 36 , thereby inhibiting cWnt signaling by preventing the interaction of LRP5/6 and frizzled receptors. It is now widely accepted that cWnt signaling plays a major role in driving the differentiation of osteoprogenitors 37 , and as such, Dkk-1 is a potent osteoinhibitory factor. The first association between Dkk-1 and osteolytic malignant bone disease was demonstrated in MM 12 then for juvenile OS 10 and skeletal metastases of breast, prostate and lung cancers 13,14,[38][39][40] . Dkk-1 was subsequently shown to block the anabolic axis of bone turnover resulting in the premise that inhibition of Dkk-1 may slow the progression of MBD by inhibiting the development of osteolytic lesions. Initial strategies for the blockade of Dkk-1 began with antibodies, and these showed promise in MM models 18  suggesting that Dkk-1 blockade might be contributing to depletion of the tumor stroma. To explore this hypothesis more closely, the ESTIMATE platform was employed to assess tumor purity and it was apparent that the DkkMo and combination groups exhibited a greater level of predicted purity as compared to the no treatment and DRB groups [Fig6G]. It has recently been reported that the calculated purity of OS tumors, and thus a relatively low proportion of stroma, is inversely proportional to the potential for epithelial to mesenchymal transition (EMT), a process which is strongly related to disease progression and poorer prognosis 57 . Therefore, the data indicated that DkkMo may also act to reduce tumor progression through depletion of host stromal elements that contribute to EMT.

20
Small interfering RNA has been employed to block Dkk-1 and exhibit neuroprotection in a model of intracerebral hemorrhage 58 , inhibit inflammation in a model of rheumatoid arthritis 59 , and in a model of hormone deficiency induced bone loss 60

Quantitative RT-PCR (qRTPCR):
The High Pure RNA isolation kit (Roche Diagnostics) was used for total RNA extraction from the cells. Copy DNA synthesis was performed using Superscript III kit (Life Technologies). TaqMan gene expression assays (Applied Biosystems) were used to carry out qRTPCR. APDH, MRPL19 and RPS18 were selected as reference genes [62][63][64][65] . Reference genes were combined using geometric averaging as single internal control gene 63 . Fold changes were calculated using the 2 −ΔΔCT method 66 .

Micro-CT (µCT) scanning:
Once euthanized, mice were fixed by trans-cardiac perfusion. All samples (n=10 each group) were scanned using a SkysScan1275 system with the filtered (1.5 mm aluminum) beam set to 40 kV, 250 µA and image capture set to 11 µm resolution.
Throughout the study, smoothing and beam hardening were fixed at 2% (smoothing kernel gaussian) and 25%, respectively. Ring artifact reduction and misalignment correction were adjusted manually to minimize scan artifacts. The dynamic range was set to between -1000 and 7519.5 HU for all reconstructions. Bone loss was calculated by comparison of volumes and bone 23 mineral densities with the contralateral side. Bone mineral density (BMD) was measured using the attenuation coefficient method with calcium hydroxyapatite phantoms (Bruker) as calibrants.
The mean of three 0.1 mm 3 regions corresponding to 25%, 37% and 50% of the total length of the bones beginning from the proximal end was measured. Half of the specimens were also contrast stained with iodine/potassium iodide (IKI) for two weeks 71  superimposed to contralateral bone model using a rigid algorithm to at least 5% accuracy.
Models were further trimmed to the same anatomical landmarks. Model to model distances were calculated with the signed closest point approach. Bone deformation measurements were generated by quantifying the differences in coordinates between corresponding voxels in scans of the tumor-bearing versus contralateral images. The deformation analysis was done by quantile functional regression for regression analyses of distributions 73 . In short, quantlet, a union set of elements was selected from the sample distribution to represent the whole distribution by lasso regression and cross-validated concordance. The quantlet from each sample was then fit to a quantlet space model using a Bayesian modeling approach. Distribution plots were analyzed for differences between one another using the using the Quantile Function on Scalar Regression

Analysis for Distributional Data method 74
High throughput sequencing: For proliferative cultures, MOSJ-Dkk1 cells were seeded at 500 cells per cm 2 in 175 cm 2 flasks and allowed to enter the logarithmic stage of growth after 6 days of culture with changes of media every 2 days. Groups (n=3) were no morpholino, 5 µM DkkMo, 5 µM scrambled morpholino control (scrMo) which was added 24 h after seeding. For the nutritional stress cohort, cells were seeded and cultured in the same manner, but exposed to

Conflict of Interest:
The authors declare no competing interests.

Data availability statement:
Until archive is completed, all raw data is available to reviewers upon request.