*3.2. Genetic Characterization of Chondrosarcoma Cell Lines*

Sequencing analysis of common mutations in chondrosarcoma identified point mutations in *IDH1* (p.R132L) in CDS01 cells and *IDH2* (p.R172G) in CDS17 and T-CDS17 cells which were also detected in the corresponding patient tumor samples. Otherwise, CDS06 cells did not show any *IDH1* or *IDH2* mutations. Analysis of *TP53* (exons 4 and 6) revealed the presence of nonsynonymous homozygous in all cell lines. The single nucleotide variants (SNV) p.P72R was found in all cell lines and also in CDS06 and CDS11 patient samples, meanwhile the mutation p.S215R was only found in CDS17 and T-CDS17 cell lines but not in the corresponding patient sample (Figure 2A,C). Additional analysis of hot spot mutations in Phosphatidylinositol-4,5-Bisphosphate 3-Kinase Catalytic Subunit Alpha (*PI3KCA*) showed no alterations in any of the cell lines. Finally, copy number analysis showed a significant gain of *MDM2* in CDS17 and T-CDS17 cells which was not detected in the tumor sample and a homozygous deletion of *CDKN2A* (exon 3) in CDS11 cells and matched patient sample (Figure 2B,C).

#### *3.3. WES of Chondrosarcoma Cell Lines and Clonal Evolution after In Vitro and In Vivo Growth*

To better characterize the genomic alterations present in these cells lines, we performed WES in the CDS17 line (passage 14), its xenograft-derived line T-CDS17 (passage 5) and their matched normal (non-tumoral) and tumor patient samples. The average nucleotide coverage in WES studies was approximately 115X, being selected for further analyses only the variants presenting more than 15 reads. Data from WES analysis of tumor the sample were compared to that of normal tissue DNA to exclude germline alterations. Tumor and CDS17 samples showed similar number of somatic mutations (123 and 121 respectively) whereas the T-CDS17 cell line displayed a slightly higher number of mutations (169), corresponding most of them to SNV in the three samples (Figure 3A). All samples displayed a similar profile of SNV transitions and transvertions (Figure 3B and Figure S2A). Similar to other type of tumors, C > T and G > A transitions were the most common mutations found in all samples (Figure 3B). To analyze the genomic evolution of tumor cells after in vitro/in vivo growth adaptation, we used variant allele frequency data of tumors, CDS17 and T-CDS17 samples to delineate the different clonal populations in each sample using the PhyloWGS, and FishPlot software. This analysis retrieves 14 clusters which evolved among samples (Figure 3C,D). The tumor sample contains nine clusters presenting cellular prevalence values higher than 0.05. Among them, cluster 1 is the one including a higher number of SNVs and must be the founder clone since the set of mutations that contain is present in virtually all the tumors cells (cellular prevalence equal to 1) and the other clones seem to derive from it. Notably, the cellular prevalence of cluster 1 is maintained in CDS17 and T-CDS17 samples, thus suggesting that most variants, including driver mutations, are kept by in the cell lines. In addition, cluster 7 remained unchanged in a small proportion of cells in all samples. On the other hand, Clusters

2–4 were positively selected while Clusters 10–14 almost disappeared during the adaptation to in vitro culture, as seen by their variation in cellular prevalence in the CDS17 line. Moreover, Clusters 5 and 6 emerged in the cell line T-CDS17 with a cellular prevalence of 0.25 and 0.15 respectively and were likely acquired during the in vivo growth of tumor cells in immunodeficient mice (Figure 3C,D).


**Figure 2.** Genetic characterization of chondrosarcoma cell lines. (**A**) Sanger sequencing chromatograms showing mutations (black arrows) in *IDH1*, *IDH2*, and *TP53* genes present in the indicated tumors and cell lines. Reference wild type (WT) sequences are shown. (**B**) Gene copy numbers of the indicated genes were estimated by quantitative PCR on genomic DNA. Results are expressed relative to the corresponding healthy tissue sample and are the mean and standard deviation of three experiments (\*: *p* < 0.05; \*\* *p* < 0.005; two-sided Student's *t*-test). (**C**) Summary of the genetic characterization of the indicated chondrosarcoma cell lines and tumor samples. Homozygous mutations are highlighted in bold.

**Figure 3.** Clonal evolution of somatic mutations after in vitro and in vivo growth of chondrosarcoma cell lines. (**A**,**B**) Mutational burden data of a tumor patient sample and its derived cell lines CDS-17 and T-CDS17 obtained by whole exome sequencing (WES). The number and type of somatic mutations (**A**) and the profile of SNV transitions and transvertions (**B**) in each sample are shown. (**C**,**D**) Subclonal reconstruction was performed with PhyloWGS using WES data. The mean cellular prevalence estimates of mutation clusters in originating patient tumor sample and subsequent CDS17 and T-CDS17 cell lines are shown (**C**). Line widths represent the relative abundance of single nucleotide variants (SNVs) in each mutation cluster. Fish-plot representation of the different clusters in each sample is also shown (**D**). (**E**) List of non-synonymous somatic mutations detected in each cluster. Catalogue of Somatic Mutations in Cancer (COSMIC) status, DREAMgenics (DG) algorithm value, and mutation type are indicated.

To select the most relevant somatic mutations included in each group of clusters presenting similar trends we compared tumor versus normal, CDS17 versus tumor and T-CDS17 versus CDS17 samples and filter the results to select variants with non-synonymous effect on coded proteins which presented variant allele frequencies >0.035 in tumor, CDS17 or T-CDS17 samples and maximum allele frequencies <0.01 in population databases (dbSNP, ExAC, ESP, and 1000 Genomes) (Figure 3E). Using this approach, we found a group of 14 mutations included in cluster 1 which were mutated in all the samples. Among these mutations, only the c.514A > G transition (p.R172G) in *IDH2*, previously detected by Sanger sequencing (Figure 2A), was listed in the Catalogue of Somatic Mutations in Cancer (COSMIC). In addition, we also selected two different mutations in the *COL2A1* gene previously found in chondrosarcoma [5,6]. Mutations in the rest of genes were not previously described in cancer. Notably, some of these unreported variants, including SNVs in Vanin2 (*VNN2*), Calcium Voltage-Gated Channel Subunit Alpha1 D (*CACNA1D*), Melanin Concentrating Hormone Receptor 2 (*MCHR2*), Unc-5 Netrin Receptor D (*UNC5D*), or Mastermind Like Transcriptional Coactivator 2 (*MAML2*), showed high values in the 0 to 5 scale assigned by DREAMgenics (DG) value (integrated score of several predictive algorithms [34], see supplemental methods for details) used to predict deleterious mutations and, therefore, they might constitute new driver events in chondrosarcoma (Figures 3E and 4B). Another group of five mutated genes was filtered from variations included in Clusters 2–4, which become enriched in CDS17 and T-CDS17 cells. The most notably alterations in this group was the mutations p.S215R and p.P72R in *TP53*, previously detected by Sanger sequencing (Figure 2A), which was present in both alleles due to a copy-neutral (CN) loss of heterozygosity (LOH) event in chromosome 17 as detailed below. Most relevant variations emerging in T-CDS17 cells (Clusters 5 and 6) included mutations in 20 genes. Among them, a SNV in F-Box and WD Repeat Domain Containing 5 (*FBXW5*) is the only variation previously reported in COSMIC in tumor types different to chondrosarcoma (Figure 3E). Also, the set of variants of the tumor sample that were lost in CDS17 and T-CDS17 cells (Clusters 10–14) included a mutation in Kinesin Family Member 21A (*KIF21A*) previously reported in COSMIC for other types of tumor and other 13 unreported mutations. Finally, three mutations were filtered from variations contained in Clusters 8 and 9, which appear in a small fraction of CDS17 cells and disappeared again in T-CDS17 cells (Figure 3E). Besides the above described changes in somatic mutations, a major consequence of in vitro and in vivo growth of tumor cells was the emergence of structural alterations and copy number variants (CNV) leading to numerous LOH events in many mutations of CDS17 and T-CDS17 cells (Figure 4A,B). Most relevant structural alterations detected in CDS17 and T-CDS17 cells included a CN-LOH affecting chromosome 17, which would explain the homozygous mutations of the *TP53* gene (p.S215R and p.P72R) commented above. Analysis of variant frequencies in this chromosome showed that the CN-LOH affected the whole chromosome and is detected in virtually all cells, as indicated by the disappearance of almost all intermediate frequencies in CDS17 and T-CDS17 cells (Figure 4B,C). Similar CN-LOH events were detected in chromosome 16, although in this case the structural variation affects to only a subset of the cells, as seen by the shift in the intermediate frequencies of variants in CDS17 and T-CDS-17 cells as compared to normal and tumor samples. Noteworthy, the intermediate frequencies shift also indicated that the subpopulation presenting the CN-LOH in chromosome 16 increased in T-CDS17 as compared to CDS-17 cells (Figure 4B,C). Besides these CN variations, other LOH events were due to CNV in several chromosomes. Thus, a copy of chromosome 18 was lost in a subpopulation of CDS-17 cells and in the whole population of non-synonymous mutations undergoing LOH in each sample. COSMIC status, DG algorithm value, and mutation type are indicated.

After filtering LOH events using the criteria described above for somatic variants and including also other LOH variants with a recurrence in the COSMIC database higher than 10, we selected 24 mutations that underwent LOH in CDS17, 17 LOH events which emerged in T-CDS-17 and 5 mutations which were detected as LOH variations in CDS17 but not in T-CDS17 (Figure 4D). Altogether, these data indicate that these cell lines kept the most relevant driver mutations present in the founder clone of the tumor sample. In addition, the adaptation of tumor cells to in vitro cell culture and in vivo growth is accompanied by the gain/loss of additional point mutations and structural variants affecting different subclones of the cell lines.

**Figure 4.** Genomic structural alterations after in vitro and in vivo growth of chondrosarcoma cell lines. (**A**) Number of LOH (total number and distribution between chromosomes) calculated from WES data of tumor, CDS17, and T-CDS17 samples. (**B**) Circular representation of the four samples sequenced through WES. Circles display from the inside outwards: (ring 1) chromosome ideogram, highlighting relevant previously reported (purple) or unreported (grey) somatic variants shared by tumor tissue, CDS17 and T-CDS17, as well as a TP53 homozygous mutation shared by CDS17 and T-CDS17 but not tumor tissue (orange); (ring 2) chromosome copy number (CN) of each sample based on normalized read counts; (rings 3–6) variant frequencies of common polymorphic positions (minimum allele frequency ≥0.01 in at least one major dbSNP population) in each sample. (**C**) Analysis of variant frequencies (left panels) and CN (right panels) extracted from (**B**) in the indicated chromosomes. Similar representations for other chromosomes are show in Figure S3B. (**D**) List of T-CDS17 cells. Conversely, a copy of chromosome 20 was gained in subset of CDS17 cells and in the entire population of T-CDS17 cells (Figure 4B,C). Finally, other copy number variations (CNVs) and/or shifts in variant frequencies affecting subpopulations of CDS-17 and/or T-CDS17 cells, was also detected in other chromosomes like 1, 2, 4, 5, 6, or 7 (Figure 4B and Figure S2B).

#### *3.4. Analysis of CSC Subpopulations in New Chondrosarcoma Cell Lines*

All the cell lines (CDS06, CDS11, CDS17, and T-CDS17) could be cultured as tumorspheres for at least two passages showing sphere forming frequencies between 0.20% and 0.28% in the first passage and between 0.1% and 0.16% in the second passage (Figure 5A and Table 1). Regarding the expression of CSC-related factors, CDS17 and T-CDS17 cells displayed high/medium levels of SOX2, while we could not detect SOX2 expression in CDS06 or CDS11 lines (Figure 5B). A similar pattern of expression was detected for ALDH1A1, while all the cell lines expressed ALDH1A3 (Figure 5B). According to the important contribution of these ALDH1A3 isoform to the Aldefluor activity [13], we detected Aldefluor positive cells in all lines in percentages ranging from 2.94% to 14.5% of the cells (Figure 5C). As previously shown in other sarcoma models [13], tumorsphere cultures of the CDS11 cell line are enriched in Aldefluor activity (Figure 5D). Furthermore, to confirm that Aldefluor could be used as a bone fide CSC marker in chondrosarcoma cell lines, we sorted T-CDS17 cells displaying high (ALDHhigh) and low (ALDHlow) Aldefluor activity (Figure 5E) and inoculated several low cellular doses (LDA assays) subcutaneously into immunodeficient mice. We found low incidences of tumor growth after the inoculation of these cellular doses (we detected tumors in 4 out of 30 mice). In any case, ALDHlow cells developed only one tumor in the series where the higher number of cells was inoculated, whereas ALDHhigh cells were able to develop three tumors, one per series. Using ELDA software to calculate the tumor initiation frequency of each population, we found that the ALDHhigh subpopulation is three times more enriched in CSCs than the ALDHlow subpopulation (Figure 5F). Although, due to relatively low tumorigenicity of these cells when injected at low cellular doses, these results do not reach statistical significance, they clearly suggest that ALDH activity could be a suitable stem cell marker in these cells.

#### *3.5. Invasion Ability of Chondrosarcoma Cell Lines*

Using live cell time-lapse microscopy we found that spheroids of CDS11, CDS17, and T-CDS17—but not CDS06 cells—were able to invade 3D collagen matrices (Figure 6A,B). In accordance with its enhanced aggressiveness, T-CDS17 also displayed a significantly increased invasive potential when compared to the parental CDS17 cell line.

Deregulated SRC/FAK (Steroid receptor coactivator/Focal adhesion kinase) signaling is related to enhanced migration and invasion in many types of tumors and we previously found that several sarcoma models may invade through a mechanism depending on SRC/FAK signaling [30]. Here, we found that both the SRC inhibitor dasatinib or the FAK inhibitor PF-573228 were able to dose-dependently inhibit cell invasion, thus indicating that this mechanism is also mediating invasion in chondrosarcoma cell lines (Figure 6C–F).

#### **4. Discussion**

Chondrosarcomas are inherently resistant to conventional treatments and a range of new therapies aimed to target specific alterations are being currently tested [9,11,12]. Among them, the use of IDH inhibitors have not proved preclinical anti-tumor activity [49] and clinical trials including chondrosarcoma patients have not yet reported positive results [12].

Other therapeutic strategies that are being tested at clinical level include the targeting of signaling pathways controlled by hedgehog, SRC or PI3K/AKT/mTOR, as well as histone deacetylase inhibitors or anti-angiogenic agents, with only a few of them reporting partially encouraging results [12,50,51]. Altogether, there is an urgent need for more research aimed to find and test new therapies for advanced or unresectable chondrosarcomas.

**Figure 5.** Characterization of cancer stem cell (CSC) subpopulations in chondrosarcoma cell lines. (**A**) Representative images of tumorspheres formed from the indicated cell lines in two successive passages. Scale bar = 100 μm. (**B**) Protein levels of Aldehyde Dehydrogenase 1 Family Member-A1 (ALDH1A1), -A3 (ALDH1A3), and Sex Determining Region Y-Box 2 (SOX2) in the indicated cell lines. β-actin levels were used as a loading control. (**C**) Aldefluor assay showing the activity of ALDH1 in the indicated cell lines. ALDH1 activity was blocked with the specific inhibitor N,N-diethylaminobenzaldehyde (DEAB) to establish the basal levels. (**D**) Comparison of Aldefluor activity in adherent and tumorsphere cultures of CDS11 cells. (**E**) Flow cytometry cell sorting of Aldefluor high (ALDH1high) and low (ALDH1low) populations in T-CDS17 cells. (**F**) Limiting dilution assay to evaluate tumor initiation frequency (TIF) of ALDH1high and ALDH1low T-CDS17 cells. The number of mice that grew tumors after 4 months and the total number of inoculated mice for each condition is indicated. TIF was calculated using ELDA software.

**Figure 6.** Invasive ability of chondrosarcoma cell lines. (**A**,**B**) Analysis of the invasive properties of the indicated cell lines using 3D spheroid invasion assays. Representative images of the 3D invading spheroids at the initial and final time points (**A**) and quantification of the invasive area (**B**) are presented. (**C**,**F**) Effect of increasing concentrations of PF-573228 (**C**,**D**) or dasatinib (**E**,**F**) on the invasive ability of CDS11 and CDS17 cells. Representative images of the 3D spheroids treated with the indicated concentrations of PF-573228 (**C**) or dasatinib (**E**) for 24 h and quantification of the invasive area after each treatment (**D**,**F**) are presented. Scale bars = 200 μm. Error bars represent the SD, and asterisks indicate statistically significant differences between the indicated series (\* *p* < 0.05; \*\* *p* < 0.01; two-sided Student's t-test). (DMSO: Dimethyl sulfoxide).

Cell lines are easy to culture, relatively inexpensive and amenable to high-throughput screening models that have guided advances in cancer research for decades. Despite limitations, such as the accumulation of new mutations after endless in vitro culture [17], international studies reported that large panels of cell lines recapitulated the most relevant alterations of original tumors and, when using a relevant number of well characterized cell lines of a given cancer subtype, they were useful models to predict anti-cancer drug sensitivity and clinical outcomes [15,52–54]. These studies have contributed to retrieve the interest in substituting veteran endless passaged cell lines and replace them with new patient-derived cell lines which should be tagged with clinical information and include genomic characterization of both the cell line and the patient samples. These cell line models would complement more sophisticated models such as organoids or patient-derived xenografts [55].

Here we present four new chondrosarcoma cell lines with related clinical information and genetic characterization of the most common alterations (mutations in *IDH1*, *IDH2*, *TP53*, and *PI3KCA* and copy number variants in *MDM2* and *CDKN2*). All the alterations detected were also found in the original tumors with the exception of the *TP53* mutations and *MDM2* amplification found in CDS17 and T-CDS17 cell lines but not in tumor samples. This finding suggests that the loss of functional p53 could be a mechanism of adaptation to in vitro culture in chondrosarcoma cells. Our cell lines expand the panel of available chondrosarcoma cell lines (overviewed in Table S3). Of note, our study provides the first cell line (CDS06) derived from a secondary chondrosarcoma associated to a previous osteochondroma. In addition we add new dedifferentiated cchondrosarcoma cell lines with *IDH2* mutations to only one reported so far. Interestingly, none of previously published dedifferentiated lines described mutations in *IDH1* (Table S3).

Relevant for possible future applications in cancer research, three of the cell lines (CDS11, CDS17, and T-CDS17) were able to initiate tumors in vivo resembling the histology of the patient sample after inoculation in heterotopic and/or orthotopic sites. Some of these cell lines were also able to invade 3D matrices and all of them showed CSC-related features such as the ability to grow as tumorspheres or the presence of subpopulations with ALDHhigh activity. Related to this, it has been previously shown that sarcoma cells increased their stemness and tumorigenic potential after being grown in mice [13,56]. Therefore cell line/xenograft line tandems, like the one formed by CDS17 and T-CDS17, constitute valuable models for studying/tracking cancer stem cells subpopulations during tumor progression [13]. These studies point ALDH1 as a relevant CSC-associated factor in different types of sarcoma [13,56]. Similarly, we found that the T-CDS17 cell line showed increased ALDH1 expression and activity, enhanced invasive ability and increased in vivo tumorigenic potential than its parental CDS17 cell line. Moreover, our results also suggest a role for ALDH1 as CSC marker in chondrosarcoma.

Our work also includes a WES analysis of CDS17 and T-CDS17 cells lines together with normal and tumor samples from the patient. This is the first time that a chondrosarcoma cell line and matched patient samples include such a level of genomic characterization. This analysis allowed us to know how the cell lines resemble the genomic diversity of the original tumor and also to track the genomic evolution of tumor cells during in vitro and in vivo growth. We found that the putative founder clone, that including a higher number of mutations and is present in all tumor cells, remains unaltered after in vitro cell culture (CDS17 cells) and in vivo growth (T-CDS17 cells). This clone includes previously known mutations, such as that previously found by Sanger sequencing in *IDH2* (R172G) [2,4] or mutations in *COL2A1* (G939Wfs\*5 and P668Lfs\*120) [5,6], as well as other unreported non-synonymous mutations presenting high scores in impact prediction algorithms, such as *VNN2*, *CACNA1D*, *MCHR2*, *UNC5D*, or *MAML2*, which possibly contribute to chondrosarcoma progression must be studied in detail.

Other mutations affecting different subclones appear or disappear during the adaptation of cells to in vitro/in vivo growth. Although not widely described for sarcomas, this genomic drift is in line with previous studies in other cancer types [57]. An important phenomenon related to the adaptation to growth conditions is the emergence of structural alterations in several chromosomes in CDS17 and T-CDS17 cells. The most relevant change was the CN-LOH affecting the chromosome 17 and responsible for appearance of homozygous mutations in *TP53*. Importantly, alterations in *TP53* were associated with aggressive behavior of chondrosarcomas [58] and similar LOH affecting chromosome 17 was detected in high grade chondrosarcomas [59]. Therefore, despite not being present in the original tumor, the CN-LOH in chromosome 17 detected in CDS17 and T-CDS17 resembles a naturally occurring mechanism for increasing aggressiveness in chondrosarcomas. Other structural variants, such as those detected chromosomes like 16, 18, or 20, affected a sequentially increased cell population in CDS17 and T-CDS17 cells. Given that T-CDS17 cells are more aggressive tan CDS17 cells, some of these structural alterations could be involved in the gain of malignancy in T-CDS17 cells. Whether

these alterations occur also in patients during tumor progression or are only due to the adaptation of tumor cells to grow ex-vivo remains to be studied.

Although we could not be completely sure that all the genetic differences observed between CDS17 and T-CDS17 cell lines were due to the in vivo growth in mice and not to the ex-vivo expansion of T-CDS17, the fact that the genomic drift previously observed in patient-derived cell lines were mainly restricted to the first few passages [57] when the adaptation of cells to the in vitro growth conditions occurs, suggests that the new set of genetic alterations detected in T-CDS17 most likely emerged during the in vivo growth phase.
