Ultrasonography of Osteosarcoma in Dogs
Sonography Atlas of Dogs

Background

Osteosarcoma (OSA) is the most common primary bone tumor of dogs with >8000 cases diagnosed annually in the United States of America [1]. Despite aggressive treatment incorporating surgery, radiation therapy, and/or chemotherapy, survival times remain poor. Over 90% of canine patients undergoing standard-of-care treatment, without clinically detectable metastasis at time of diagnosis, will ultimately succumb to metastatic disease [1,2]. Overall reported median survival times are 235-366 days, and 1- and 2-year survival rates range from 33-65% and 16-28%, respectively [1-4]. The development of effective tools to diagnose and treat OSA is essential for mitigating the negative impact of this disease on both canine patients and their owners.

The extracellular surface of a cancerous cell includes a multifarious array of proteins that are critical to processes such as cell proliferation, migration, adhesion, and inter-cellular communication. Collectively, this group of proteins can be referred to as the surface-exposed proteome (SEP). Abnormalities in the SEP are central to the biologic behavior of cancer [5]. Recognized abnormalities in cancerous cells include changes in the expression level of particular proteins, the presence of unusual protein isoforms, and alterations in post-translational modification patterns (e.g. phosphorylation or glycosylation) [5,6].

Despite OSA’s considerable clinical impact, a comparative analysis of the SEP between canine osteosarcoma and normal osteoblast cells has not been described in the veterinary literature. Isolated reports of specific abnormalities in canine OSA are limited to focused evaluations of specific proteins, often prompted by discoveries in other neoplasms or species. Examples of such proteins include: survivin [7], cathepsin K [8], Met [9-11], EGFR [9], Ron [9], p53 [12,13], HER-2 [14], HGF [10,11], IGF-1 and IGF-1R [15], MMP-2 and MMP-9 [16,17], and ezrin [18]. A more global analysis of the SEP differences between OSA and normal osteoblast cells holds the potential of identifying unique and previously unrecognized features of OSA that can be exploited in treatment, diagnostic, and/or prognostic capacities [5,6,19-23].

While there are many different methods available for enriching plasma membrane proteins, labeling surface proteins with biotin is a popular approach because of the availability of membrane-impermeable biotinylation reagents and the strong association between biotin and streptavidin, which is used for affinity purification of biotin-labeled molecules [24]. These two attributes make it feasible to selectively label surface proteins, and then use denaturing conditions to extract total protein from labeled cells and affinity purify the biotinylated fraction. The purified surface proteins can then be enzymatically digested with trypsin and the resulting peptides identified by mass spectrometry (MS). MS-based proteomics has been applied to numerous cancer biomarker discovery projects and enables protein-level investigations reflecting many of the biological processes relevant to cancer growth and invasion [20,21,25,26].

The specific aim of our study was to define a profile of proteins that compose the SEP of two validated canine OSA cell lines and a normal canine osteoblast cell line utilizing the aforementioned biotinylation/streptavidin system to selectively label, purify, and identify surface-exposed proteins by LC-MS/MS analysis. Additionally, we sought to validate a subset of our MS-based observations via quantitative real-time PCR (qRT-PCR) and antibody-based techniques. Our hypothesis was that MS would detect differences in the SEP composition between the OSA and the cultured normal osteoblast cells.

Results

Identification of surface-exposed proteins

The biotin-labeled surface proteins from each cell line were extracted and visualized by SDS-PAGE and Western blot (Figure 1). The gross comparison of biotinylation patterns indicated that the complement of surface-exposed proteins from the POS and HMPOS cell lines appeared similar, while the surface proteome of the CnOb cell line appeared to be markedly different. Following the affinity purification and proteolysis of the biotinylated proteins, analysis of the peptides by shotgun MS identified a total of 133 putative surface proteins when outputs from all samples and replicates were combined (Additional file 1: Table S1). Protein identifications were largely consistent between biological replicates (Additional file 2: Table S2). Based on their potential relevance to cancer biology, a subset of the identified proteins was selected for inclusion in Table 1, with their relative spectral counts expressed as a percentage of total spectra. Several proteins that are known to be associated with the cell surface and/or interact with the extracellular matrix were detected in all three cell lines, such as Fibronectin, Vitronectin, CYR61 and Annexin A2 [27-30]. Several additional surface proteins were found to be detected in abundance in one cell line, but poorly detected or absent in the others. For example, peptides originating from Thrombospondin-1 were observed in abundance from CnOb samples, but were not observed in any of the replicates from the POS or HMPOS cell lines. Similarly, both Chondroitin Sulfate Proteoglycan 4 (CSPG4) and CD109 were detected from POS samples, but were not detected in the other samples.

qRT-PCR

Eleven of the proteins that were detected by MS were selected for secondary analysis of gene expression by qRT-PCR, the results of which are summarized in Figure 2. These eleven proteins were selected on the basis of their relevance to cancer biology and the availability of antibodies that were potentially compatible with canine targets. Active transcription of all of the selected genes was detected by qRT-PCR, indicating that all of these genes are indeed expressed in the cell lines used in this study. Expression was normalized to GAPDH and the biological replicates were generally in close agreement between samples, with few minor variations (Additional file 3: Table S3).

Western blot

Following validation and quantification of gene expression by qRT-PCR, five proteins were selected for further validation by Western blot: Thrombospondin-1, CYR61, CD44, Notch2 and Plexin B2. Due to the dearth of validated, canine-specific antibodies, these targets were selected based on the availability of antibodies that were likely to be cross-reactive with canine proteins. The antibodies against Notch2 and Plexin B2 were found to have poor target specificity towards the canine proteins in our samples and were excluded from the final figure. Western blots of whole cell lysates from all three cell lines was effective for Thrombospondin-1, CYR61 and CD44, and indicated that all three proteins were present in each cell line, albeit with differential abundance (Figure 3).

Immunocytochemistry

To assess surface-localization of the selected proteins, an on-cell Western blot approach was used to visualize and quantify the proteins on the surface of live cells. Of the five experimental antibodies tested in this study, only the anti-CD44 antibody was validated for either immunocytochemistry or flow cytometry applications, and only with murine cells. Not surprisingly, only the anti-CD44 antibody was observed to bind, at detectable levels, to the cell surface of the cell lines used in this study. The fluorescent signal from anti-CD44 was quantified on a fluorescent scanner and normalized to the approximate number of cells, as measured by the signal intensity from the nuclear stain Hoecsht 33342. This semi-quantitative ICC analysis indicated that all three cell lines expressed CD44 on their surface, although it was much less abundant on the surface of CnOb cells compared to POS or HMPOS cells (Figure 4). The complete ICC results for all of the tested antibodies are summarized in Additional file 4.

Discussion

The present study successfully employed a method of biotinylation/streptavidin enrichment with MS-based peptide sequencing to selectively label, purify, and identify surface exposed proteins from two validated canine OSA cell lines and a commercially available normal osteoblast line. A total of 133 putative surface-exposed proteins were identified when outcome from all cell lines was combined. Significant differences in SEP composition, and many similarities, were observed between the canine OSA cell lines (POS, HMPOS) and the normal canine osteoblast cell line (CnOb). Based on these observations, we conclude that MS can effectively detect SEP differences between cultured canine OSA and normal osteoblast cell lines.

When evaluating the list of 133 proteins identified via these methods, it is important to bear in mind that confident quantification of protein levels is not possible with this methodology because relative quantities can become skewed during the enrichment process, depending on the protein. Furthermore, this analysis invariably represents only a fraction of the total SEP; as a comprehensive profile would require the use of complementary sample processing methods, standards, and controls. Despite these limitations, the primary benefit of MS-based proteomic techniques, such as the one described in this report, is the ability to identify numerous previously unrecognized SEP differences between samples in a high-throughput fashion. This latitudinous analysis of the SEP can highlight novel targets for future diagnostic, prognostic and therapeutic methods [5,6,19].

MS-based proteomics is a relatively new tool in canine OSA research, and due to the limited information on this specific disease, we sought to validate a subset of our MS observations via secondary methods. These included qRT-PCR, Western blot, and immunocytochemistry. The results from these secondary tests were broadly confirmatory, but they occasionally contradictory, and highlighted some important limitations of the MS data sets. These results are summarized for a selected group of proteins in Table 2. For example, a negative result for an antigen in the MS data was not necessarily indicative that the target was absent, which was illustrated by our confirmation of CD44. Overall, the limited scope of this study means that the resulting data is inherently incomplete. However, these initial results also indicate that further improvements in sample processing methods, mass spectrometry technologies, and canine-specific antibodies will improve the feasibility and utility of this type of analysis.

Our initial supportive experiment was qRT-PCR analysis of 11 of the 133 MS-identified surface-proteins. These 11 proteins were selected based on their known roles in cancer biology and apparent differences in abundance, as measured by MS spectral counts. Our analysis revealed that transcripts for all 11 genes were detected in each of the 3 cell lines, indicating that each of these genes were indeed expressed, and that each protein was potentially produced by each cell line. Protein quantification using spectral counting (label free approach) is considered to be semi-quantitative and the accuracy is based on the length of the protein, reproducibility of the sample preparation techniques, and LC separation [31]. In this study, because affinity enrichment was incorporated into the sample preparation, the relative abundance of the proteins determined by spectral counting can be misleading. Bearing this limitation in mind, the expression levels of several of the genes matched their apparent abundance at the cell surface, as measured by MS. Most notable among these was Thrombospondin-1, which was detected in unique abundance from CnOb samples, and whose transcript was also expressed at much higher levels in CnOb cells than in POS or HMPOS cells. At the same time, the apparent protein surface abundance did not closely match gene expression data for several of the targets. These differences may arise from differences in post-transcriptional regulation and/or protein trafficking to the cell surface, or they may be artifacts of the experimental method. While qRT-PCR demonstrates which genes are expressed and is an indicator of which proteins are likely to be more or less abundant, qRT-PCR is not a direct indicator of the abundance and localization of the corresponding proteins.

Western blot analysis demonstrated that the proteins Thrombospondin-1, CYR61 and CD44 were present at varying levels in all 3 cell lines. These results are compatible with the qRT-PCR results, indicating that these genes are both expressed and translated into protein in the cell lines studied. Again, the results were most consistent for Thrombospondin-1 protein, which was observed by Western Blot to be most abundant in the CnOb sample. Our MS analysis did not detect thombospondin-1 in the OSA cell lines, although the qRT-PCR and Western Blot results do indicate that this gene is transcribed and translated in the OSA cell lines, albeit at much lower levels than in CnOb. Additionally, CD44 was detected by qRT-PCR and Western blot in all three cell lines, but was not observed by MS in the CnOb cell line. The Western blot analysis represents all the proteins in the cells (whole cell lysate) whereas our MS-based SEP analysis is designed to specifically target the cell surface proteins. Therefore, the observed differences between Western blot results and MS findings may be due to changes in the protein localization with the cells, or they may represent limitations of our methods. Importantly, none of the proteins identified by MS were found to be absent in any cell line by Western blot (i.e. MS did not spuriously report presence of the three proteins, as verified by Western blot).

Immunocytochemistry represents the most direct method for confirming the surface localization of the proteins in question. However, the lack of validated antibodies for use with canine cells limits the population of proteins that could be confirmed using this technique. While all five of our antibodies were tested for potential reactivity, only one (αCD44) was found to bind to the surface of any of the cell lines, at a detectable level. This was result was not surprising because only the αCD44 was validated for flow cytometry applications. We observed positive CD44 staining for all of the cell lines, although the signal was much greater in the OSA lines compared to the CnOb cells. These results are largely consistent with our MS-based SEP analysis and are compatible with our Western blot observations. A previous study, using methods other than MS, has also demonstrated presence of CD44 on the cell surface of cultured canine OSA cells [32]. Furthermore, the CD44 immunocytochemistry findings suggest gene expression levels (assessed by qRT-PCR) and presence of CD44 in the whole cell lysate (assessed by Western blot) may not be indicative of concentration of the protein in question at the cellular surface—an observation that supports the selectivity of the biotin/streptavidin surface-protein enrichment technique employed in the present study.

Although the 133 proteins identified in this study represent a significant contribution to the study of canine OSA biology, they do not represent the entire SEP of these cells. A more comprehensive survey of the SEP composition could be achieved with the addition of complementary sample processing methods. For example, our method used amine-reactive biotinylation followed by Trypsin digestion of the recovered proteins to generate peptides for analysis. A more thorough SEP profile might incorporate an alternative biotinylation reagent (e.g. biotin hydrazide) [33] and/or protease (e.g. Glu-C) [34,35]. The addition of these methods would generate complementary data sets, ultimately increasing the depth, breadth, and accuracy of the analysis. Also, simply increasing the physical amount of sample analyzed for each cell line would be expected to yield more peptides for detection via MS, which would in turn increase the detection rate for proteins that exist in low quantities on the cell surface.

Share this article / Teilen Sie diesen Artikel