Despite modern multimodality therapies, approximately 20% of patients with osteosarcoma show radiographically demonstrable metastatic disease at initial diagnosis, whereas most of the remainder of patients are assumed to harbor subclinical micrometastases. Unfortunately, 30% to 40% of patients eventually develop metastatic disease despite conventional treatment. For osteosarcoma, relatively little is known regarding the various interactions between host and tumor cells governing growth and progression in vivo. Murine models of osteosarcoma may provide greater insight into these interactions. Moreover, there is the potential to test various therapeutic agents and observe the in vivo response, not only at the primary site, but also in terms of the development of metastases [12]. Although a number of osteosarcoma cell lines are established in vitro, there is a paucity of reliable and reproducible in vivo animal models resembling aspects of the human condition at the temporal, physiologic, and histopathologic level [12].

Jia et al. [23] identified two cell lines derived from the well-characterized SaOS2 parent for their respective low (LM2) and high (LM7) potential for generating pulmonary metastases when inoculated intravenously into nude mice [23]. Differences in gene expression between these two cell lines may underlie the differential potential for developing pulmonary lesions. However, correlative gene expression profiles from these cell lines have yet to be characterized either in vitro or in vivo.

In these experiments, we assessed the profile of differentially expressed genes in vitro and in vivo that may be associated with the evolution of tumor metastases from a primary lesion and we established a murine xenograft model of osteosarcoma pulmonary metastasis using SaOS2 LM2 and SaOS2 LM7 cell lines. We hypothesized differences in the pattern of gene expression exist between LM2 and LM7 cells would explain the previously demonstrated differential in metastatic potential [23]. Secondly, we hypothesized the profiles of gene expression in LM2 and LM7 cells cultivated in vitro would resemble those of the respective cell lines recovered 5 weeks after implantation into tibial metaphysis of nude mice.

We studied differential gene expression between two human SaOS2-derived osteosarcoma cell lines with differing metastatic potential (LM2 and LM7) both in vitro and in vivo. cDNA microarray was used to establish a gene expression profile comparison between weakly (LM2) and highly (LM7) metastatic osteosarcoma cells in vitro. We then orthotopically inoculated 1 × 10⁶ LM2 or LM7 cells into the tibial metaphysis of four mice for each cell line, allowed 5 weeks for tumor establishment and metastasis, and extracted RNA from the inoculated tissues. From the array data, we selected several genes showing differential expression between the LM2 and LM7 cell lines in vitro and used reverse transcriptase polymerase chain reaction (RT-PCR) to assay their expression in vivo.

The LM2 (low-potential) and LM7 (high-potential) SaOS2-derived cell lines were a gift provided by Dr Eugenie Kleinerman at the M.D. Anderson Cancer Center (Houston, TX) [23] and maintained in minimum essential medium supplemented with 10% fetal calf serum, 1% nonessential amino acids, 1 mmol/L sodium pyruvate, 50 IU/mL penicillin, and 50 μg/mL streptomycin (MediaTech, Manassas, VA).

Cells were trypsinized at approximately 80% confluence and passaged 1:4. At each of three consecutive passages, approximately 2.5 × 10⁶ cells were collected, washed in phosphate-buffered saline (pH 7.4), and pelleted by centrifugation. Cell pellets were then lysed with RLT buffer with 0.1% β-mercaptoethanol and total RNA was extracted using the RNeasy^® kit with on-column DNase I digestion (Qiagen, Valencia, CA). Samples of total RNA were assayed for RNA quality and concentration by microcapillary electrophoresis and ultraviolet absorbance spectroscopy (RNA Pico; Agilent Technologies, Santa Clara, CA).

For microarray studies, 1 μg total RNA from each of three consecutive passages was pooled for each cell line. First-strand template cDNA was generated by reverse-transcription with an oligo(dT)₁₅ primer coupled to a T7 RNA polymerase recognition sequence (Applied Biosystems, Framingham, MA). Residual single-stranded RNA was removed by RNase H digestion and 200 ng of the cDNA template was used for in vitro transcription with biotinylated CTP and UTP nucleosides to produce a cRNA template. After purification and quantification, 15 μg of the biotinylated cRNA was fragmented by hydrolysis, producing 35 to 200 nucleotide segments. Known concentrations of several positive control genes (50 pmol/L Oligo B2, 1.5, 5, 25, and 100 pmol/L of Escherichia coli bioB, bioC, bioD, cre) were also added to the cRNA template before hybridization to facilitate data standardization. Samples were then hybridized at 45° C for 16 hours to the GeneChip^® Human Genome U133 Plus 2.0 array (Affymetrix, Santa Clara, CA) according to the manufacturers recommended protocol. Fluorescence images of the gene chips were then acquired using the Agilent G2500A Gene Array Scanner and evaluated using the Affymetrix software (GeneChip Operating System, Santa Clara, CA). The overall chip intensities for each sample were scaled by linear adjustment to the same target value (500). GeneTraffic^® 2.7 (Iobion Informatics LLC, La Jolla, CA) was used to determine differential expression of genes between pooled samples using the Robust Multichip Analysis method. Differential expression was defined as a twofold or greater difference in normalized fluorescence intensity between the LM7 (numerator) and LM2 (denominator) cells. The differentially expressed genes identified from cDNA microarray comparisons were then assigned to six non-mutually exclusive metastasis-associated processes (proliferation/apoptosis, motility/cytoskeleton, invasion, adhesion, immune surveillance, and angiogenesis) using a PubMed search of the gene names as initially described by Khanna et al. [25].

The LM2 and LM7 cells lines have not yet been successfully incorporated into an animal model. After a review of the literature, we chose to inoculate mice through a tibial injection to generate a physiologic model of metastatic osteosarcoma [1, 3, 5, 9, 31]. Eight 22- to 25-g female homozygous NCr nude mice (Charles River, Wilmington, MA) were used for the xenograft experiments: four mice were orthotopically injected with LM2 cells and four mice with LM7 cells. Mice were immobilized for cell implantation and subsequent radiography by an intramuscular injection of an anesthetic cocktail containing 45 mg/kg Telazol^® (Fort Dodge Animal Health, Fort Dodge, IA) and 7.5 mg/kg xylazine (Phoenix Pharmaceuticals, St Joseph, MO). All animal protocols were performed under protocols approved by the Institutional Animal Care and Use Committee.

For xenograft experiments, LM2 or LM7 cells from Passage 6 or 11, respectively, were grown to approximately 80% confluence. Immediately before inoculation, cells were trypsinized and resuspended in growth medium at a concentration of 5 × 10⁴/μL, and 1 × 10⁶ LM2 or LM7 cells (20 μL) were inoculated percutaneously into the left tibial metaphysis as previously described [3, 5, 9, 31]. Faxitron radiographs (Faxitron X-ray Corp, Wheeling, IL) were taken immediately after tumor inoculation, 2 weeks later, and at weekly intervals thereafter and evaluated for evidence of skeletal and/or soft tissue lesions. There were no infections or perioperative complications. Thirty-five days after inoculation, the animals were euthanized by CO₂ inhalation and bilateral tibiae were harvested and rapidly frozen in liquid nitrogen for gene expression analyses. The abdominal and thoracic organs were studied grossly for evidence of soft tissue metastases, with particular attention paid to the lungs before formalin preservation for future histopathologic studies.

Osteosarcoma continues to be a fatal disease despite recent advances in research into understanding cellular and molecular biology. Approximately 80% of patients are believed to harbor metastases at the time of initial presentation. In vitro analysis of osteosarcoma cell lines has been extremely useful in identifying genetic targets in metastatic osteosarcoma. In vitro analysis, however, is inherently restricted by its nonphysiologic nature. In vivo models incorporating all the physiologic features of osteosarcoma are crucial to understanding, treating, and eventually curing metastatic osteosarcoma. We compared in vitro and in vivo gene expression of metastatic osteosarcoma and established a reliable animal model of metastatic osteosarcoma.

There are a number of limitations to the current approach in the study of metastatic osteosarcoma. Only two metastatic osteosarcoma cell lines derived from a single parent cell line were investigated in this study. However, primary osteosarcoma is a genetically complex and heterogeneous tumor in which the molecular basis of malignant transformation and metastasis remains poorly understood. Genetic variability in metastatic osteosarcoma is likely to be even greater and to vary between cell lines. Despite this, the SaOS2 is a well-established and highly studied osteosarcoma cell line. Future studies including several metastatic osteosarcoma cell lines would help confirm such genetic variability. The percutaneous approach of our orthotopic injection model is another limitation because it failed to yield radiographically demonstrable lesions at the injection site and likely disseminated the injected cells to other sites, including the contralateral tibia. However, although the establishment of a reproducible orthotopic injection model of osteosarcoma was a potential side benefit of this project at its outset, it was not the primary purpose of orthotopic injection. The primary purpose was to provide a consistent in vivo environment with which to compare the LM7 and LM2 cell lines’ gene expression. As noted earlier, further refinement of this orthotopic injection technique is necessary before this can become a standardized model. The small numbers of animals may have further compromised the less-than-complete establishment of tumor cells in the injected tibiae and diminished our ability to derive larger amounts of RNA that would allow more comprehensive molecular analysis. However, for the purposes of comparing gene expression with the most highly changed genes from the in vitro data, the amount of RNA obtained was sufficient. Finally, although RT-PCR is a valid way to confirm cDNA microarray findings, further investigation using immunostaining perhaps would have been useful in confirming our results as well.

A large number of osteosarcoma cell lines are currently listed as available from the American Tissue Culture Center. Many of these cell lines have been studied in depth in vitro resulting in considerable profiling of gene expression for human osteosarcoma. Metastatic osteosarcoma has proven more difficult to study. Several murine osteosarcoma cell lines with a higher in vivo metastatic potential currently exist, including UMR 106-01, K7M2, K12, Dunn, and LM8 [11, 14, 26]. However, human osteosarcoma cell lines are more clinically relevant to understanding clinical progression to metastasis and overall survival. Several human osteosarcoma cell lines with metastatic potential exist (U2-OS, HOS, KRIB, SaOS2, 143B), and are currently under investigation for their clinical relevance.

The SaOS2 cell line was derived by Fogh et al. [15] in 1973 from an 11-year-old Caucasian girl. Jia et al. [23] have generated a series of SaOS2-derived cell lines with varying metastatic potential. These cell lines, including the LM2 and LM7 lines used in our experiments, were cloned from lung metastases recovered from mice inoculated with SaOS2 cells by tail vein injections. Demonstration of differential gene expression between the highly metastatic LM7 derivative of SaOS2 osteosarcoma cell line and its lower metastatic potential counterpart, LM2, may provide clues as to genes important to metastases. However, in vitro data have been criticized as being overly susceptible to varying culture conditions that may invalidate the data. Hence, correlative in vivo evaluation has been suggested. To date, there have been no reports describing successful orthotopic implantation of SaOS2 LM2/LM7 cell lines.

The results of the companion in vivo study showed genes identified by microarray that were up- or downregulated in the highly metastatic LM7 cell line compared to the weakly metastatic LM2 cell line in vitro were not necessarily similarly up- or downregulated after an orthotopic injection in a mouse tibia. Although the three upregulated genes (Col11α1, IL-1R, CRLF-1) identified in our in vitro study were also upregulated in vivo, they were quantitatively less upregulated. Furthermore, the data for the downregulated genes (IL-7, MME, ADAM22) compared in the in vitro and in vivo studies were contradictory, and human homologs of several genes (CXCL14, CLU, CXCL13, PACE1, IL6R, OMD) were not detected at all in vivo.

Although some literature suggests in vitro gene expression profiling by microarray analysis does correlate with gene expression in an in vivo model [18, 33, 40], there are a number of studies that derive the opposite conclusion. When profiling breast, prostate, and glioma cells after single versus fractionated doses of radiation, Tsai et al. [42] reported genes whose expression was elevated after radiation in vitro were not elevated when these cells were exposed to radiation in vivo. Furthermore, Lund et al. [30] reported differences in gene expression in microglia when comparing cells in culture and in vivo. Finally, after finding discrepant results with their work with glioma cells in vitro and in vivo, Tatenhorst et al. [41] concluded one should be cautious in simply extrapolating data obtained in vitro to processes occurring inside the brain (in vivo). Explanations for these differences generally relate to the limitations of the rigid in vitro environmental conditions that may produce differences in gene expression with small changes in one or more numerous variables and, in some tail vein or heart injection models, to the complexity of the multistep metastatic process and regulation of its pathways. The tumor microenvironment (blood flow, hypoxia, cell cycle) in vivo is likely different enough to play a major role in the differing gene expression.

Our in vivo model was an attempt to establish an orthotopic murine model of an immortal human osteosarcoma cell line. We demonstrated by RT-PCR human GADPH, signifying presence of the injected human osteosarcoma cells, was present in some of the injected limbs. Furthermore, human GADPH was also present in two of the contralateral, noninjected limbs, suggesting the possibility of metastasis to distant bones. Some of the difficulty in establishing radiographic and gross evidence of orthotopic disease in the injected tibias may be the result of the method and site of inoculation. Previous reports have cited success in developing orthotopic osteosarcomas by percutaneous injection of SaOS2 cells into the proximal tibia [9], but success using this method with the SaOS2-derived LM2 and LM7 cell lines has not been reported. Previous reports have documented success in establishing local tumor growth after orthotopic implantation of breast cancer cells into the distal femur of mice using a small arthrotomy [1]. Therefore, it is conceivable success may be enhanced by visualization and verification of the cell inoculation into the distal femur, because this is a more common site of osteosarcoma in humans using an open arthrotomy approach. Additionally, an incubation period of greater than 5 weeks may be necessary for establishment of a radiographically evident lesion using this model.

Despite the growing body of literature describing the genetic aberrations in osteosarcoma, there still remains a relative paucity of investigations into the genetic foundation of metastatic osteosarcoma. We attempted to identify the genes responsible for allowing primary osteosarcoma cells to metastasize through a murine model of metastatic osteosarcoma. Although we observed poor correspondence of our in vitro to in vivo gene expression in this initial work, we are continuing to develop an orthotopic murine model of metastatic osteosarcoma and intend to explore the molecular and genetic foundation of metastasis in an effort to evaluate novel therapeutic strategies for patients with osteosarcoma.