IDH Mutations in Chondrosarcoma: Case Closed or Not?

Simple Summary Chondrosarcomas are cartilage tumours that often harbour a mutation in one of the isocitrate dehydrogenase (IDH) genes. IDH mutations are important drivers at the beginning of cartilage tumour development, but their role in later stages remains unclear. However, other IDH mutant tumour types do show an influence of this mutation on patient outcomes and therapies that specifically kill these IDH mutant tumour cells. Factors that could explain this discrepancy in the role of IDH mutations are differences in tumour type, elevated oncometabolite levels, the type of model used in preclinical studies (natural vs. introduced IDH mutation), and additional (epi)genetic alterations. The latter influence the downstream biological effects of an IDH mutation, and recent studies have indeed identified subgroups within IDH wildtype and mutant chondrosarcomas. Future studies should build upon these subgroups to improve the identification of effective treatments and biomarkers that predict which patients will benefit from these therapies. Abstract Chondrosarcomas are malignant cartilage-producing tumours that frequently harbour isocitrate dehydrogenase 1 and -2 (IDH) gene mutations. Several studies have confirmed that these mutations are key players in the early stages of cartilage tumour development, but their role in later stages remains ambiguous. The prognostic value of IDH mutations remains unclear and preclinical studies have not identified effective treatment modalities (in)directly targeting these mutations. In contrast, the IDH mutation status is a prognostic factor in other cancers, and IDH mutant inhibitors as well as therapeutic strategies targeting the underlying vulnerabilities induced by IDH mutations seem effective in these tumour types. This discrepancy in findings might be ascribed to a difference in tumour type, elevated D-2-hydroxyglutarate levels, and the type of in vitro model (endogenous vs. genetically modified) used in preclinical studies. Moreover, recent studies suggest that the (epi)genetic landscape in which the IDH mutation functions is an important factor to consider when investigating potential therapeutic strategies or patient outcomes. These findings imply that the dichotomy between IDH wildtype and mutant is too simplistic and additional subgroups indeed exist within chondrosarcoma. Future studies should focus on the identification, characterisation, and tailoring of treatments towards these biological subgroups within IDH wildtype and mutant chondrosarcoma.


Introduction
Chondrosarcomas are malignant cartilage-producing tumours that account for 20% of all malignant bone tumours [1,2]. Enchondromas are considered the benign precursor lesions of chondrosarcoma, but progression towards malignant tumours is rarely seen (<1%) outside the non-hereditary syndromes (i.e., Ollier disease and Maffucci syndrome) that cause multiple cartilaginous neoplasms (enchondromatosis) [3,4]. Chondrosarcomas arise predominantly in the third to sixth decades of life and can affect the long as well as the flat bones, especially the femur, humerus, pelvis, and ribs, and occasionally the spine or base of the skull. Pathological characteristics divide chondrosarcoma into several subtypes, including conventional chondrosarcoma (85%), dedifferentiated chondrosarcoma (10%), and rare subtypes that include mesenchymal, clear cell, and periosteal chondrosarcoma (5%). Based on the anatomical location, conventional chondrosarcoma can be further subdivided into central (i.e., in the medulla of the bone) and peripheral (i.e., at the surface of the bone) conventional chondrosarcoma (85% and 15%, respectively) [1,2].
Histological grading is defined by, among other factors, the mitotic count, the presence of spindle-shaped cells, cellularity, and the matrix production of the tumour, and it is the most important factor to predict overall patient survival and metastatic potential. Patients with well-differentiated tumours (i.e., atypical cartilaginous tumour (ACT) and grade I) have an overall 10-year survival rate of 88-95% and rarely show metastasis formation [1]. However, high-grade tumours (i.e., grade II and III) show increased metastatic potential (10-30% and 32-71%, respectively) and the overall 10-year survival rate of these patients is severely decreased (58-86% and 26-55%, respectively) [2]. Dedifferentiated chondrosarcoma is a high-grade subtype of chondrosarcoma with the bimorphic histological appearance of a conventional chondrosarcoma juxtaposed with a high-grade anaplastic sarcoma [5]. It has a dismal prognosis, with 5-year overall survival of only 7-24%.
The worse prognosis of both high-grade conventional and dedifferentiated chondrosarcoma can be partially ascribed to the limited number of available treatment options. Chondrosarcomas are intrinsically resistant towards chemo-and radiotherapy and targeted therapeutic options are still lacking, leaving surgery as the only curative treatment option [6]. Hence, there is an urgent need to develop novel targeted therapeutic strategies, especially for patients with metastasised and/or unresectable high-grade or dedifferentiated chondrosarcomas.
In the last decade, recurrent heterozygous hotspot mutations in the arginine residues of the isocitrate dehydrogenase 1 and −2 (IDH1 and IDH2) genes (p.R132 and p.R140/p.R172, respectively) were identified in enchondroma (87%), central conventional chondrosarcoma (~50%), and dedifferentiated chondrosarcoma (>80%) [7][8][9][10]. The high frequency of IDH1 and IDH2 (collectively referred to as IDH) mutations in benign cartilage tumours indicates that these mutations occur early in tumourigenesis, suggesting that IDH mutations have an important driver role in the formation of cartilage tumours. Indeed, the introduction of an IDH mutation induces enchondroma-like lesions in mice [11]. Furthermore, the IDH mutation or its produced oncometabolite stimulate chondrogenic differentiation while inhibiting the osteogenic differentiation of mesenchymal stem cells, which are the presumed cells of origin of cartilage tumours [12,13]. Despite their significant role in the early stages of tumour development, the prognostic value of the IDH mutation in chondrosarcoma seems controversial and (pre)clinical studies that have focused on the direct and indirect targeting of the IDH mutation have not yielded novel treatment strategies. This review provides an overview of the current knowledge of the role of IDH mutations in chondrosarcoma and highlights similarities as well as differences between tumour types that frequently harbour IDH mutations. Additionally, it will be discussed whether the IDH mutation should still be considered as a promising therapeutic target or not.

Frequency and Prognostic Value of IDH1 and IDH2 Mutations
IDH mutations are also frequently observed in other tumour types, such as acute myeloid leukaemia (AML), glioma, and cholangiocarcinoma [14]. Interestingly, the most common variant differs between the above-stated tumour types (Table 1). Cartilage tumours and cholangiocarcinoma mainly have IDH1 p.R132C variants (~60%), glioma predominantly harbours IDH1 p.R132H mutations (~90%), and AML often has IDH2 p.R140Q mutations (~40%) [15,16]. None of the variants are exclusively observed in one tumour type, suggesting that different point mutations can have a similar effect on tumourigenesis, although the level of the oncometabolite D-2-hydroxyglutarate (D-2-HG) produced by these variants differs [17][18][19]. The prognostic value of IDH mutations in these tumour types is also diverse ( Table 1), and only glioma patients have a clear favourable outcome when their tumour harbours an IDH mutation [20][21][22][23]. Studies that were performed to determine the prognostic value of IDH mutations in chondrosarcoma show contradictory results. While it was previously reported that IDH mutations do not predict outcomes [15], other studies showed either a worse [24] or better [25] prognosis for IDH mutant (IDH MUT ) chondrosarcoma patients. The three patient cohorts were similar in size (n = 70 to 80) and median age (50 to 60 years), but the chondrosarcoma subtype inclusion (conventional versus addition of dedifferentiated and mesenchymal cases) and median follow-up time (4.3 versus ≥10 years) differed, which might explain the discrepancy in results. Another factor might be the type of technique used to assign patients to the IDH MUT subgroup. For instance, Sanger sequencing is not sensitive enough to detect mutations when present in less than <30% of the sequenced PCR product, leading to false-negative results in samples with a low IDH MUT variant allele frequency or tumour cell percentage and thereby the assignment of IDH MUT patients to the IDH wildtype (IDH WT ) subgroup. Despite the lack of prognostic value, the high occurrence rate of IDH mutations in all of these tumour types suggests that they have an important role in driving tumourigenesis, already in the early stages of tumour development.

Oncogenic Activities of IDH Mutations
Both IDH enzymes function in the tricarboxylic acid (TCA) cycle, where they convert isocitrate into α-ketoglutarate (α-KG) and CO 2 . Mutated IDH enzymes acquire a neomorphic function, leading to the additional conversion of α-KG into the oncometabolite D-2-HG [39]. The IDH1 p.R132C variant is one of the most efficient D-2-HG producers, while both IDH1 p.R132H and IDH2 p.R140Q produce lower levels of the oncometabolite [17][18][19]. As certain variants are more frequently observed in specific tumour types (Table 1) [15,16], this could suggest that chondrosarcoma and cholangiocarcinoma rely on high D-2-HG levels, while glioma and AML depend on relatively lower levels of the oncometabolite.
Due to the high structural similarity between α-KG and its antagonist D-2-HG, the oncometabolite is able to competitively bind α-KG-dependent enzymes, leading to the overall inhibition of this class of enzymes [40,41]. The inhibition of α-KG-dependent enzymes leads to widespread changes in the epigenomes and metabolomes of cells and affects DNA repair and cellular growth signalling pathways ( Figure 1) [42,43]. For instance, the D-2-HG-mediated inhibition of α-KG-dependent DNA demethylases (family of TET enzymes, including TET1/2) and histone demethylases (family of Jumonji enzymes, including KDMA4A/B) leads to an overall DNA hypermethylation phenotype, as well as an aberrant histone methylation phenotype in IDH mutant tumours. IDH MUT enchondromas and chondrosarcomas are indeed characterised by a CpG island methylator phenotype (CIMP)-positive status, and DNA hypermethylation is present in primary IDH MUT chondrosarcomas [7,44,45]. The family of Jumonji enzymes is also involved in the regulation of the Mechanistic Target Of Rapamycin Kinase (mTOR) signalling pathway, as well as DNA repair via the homologous recombination pathway. Moreover, IDH MUT enzymes have a reduced ability to produce NADPH and consume high levels of NADPH to produce D-2-HG, resulting in severely reduced overall NADPH levels. This deficiency does not only cause metabolic stress but will also lead to an increase in reactive oxygen species (ROS), making IDH MUT tumours more vulnerable to DNA damage. Besides the induction of metabolic stress, IDH MUT tumours also undergo metabolic rewiring, including alterations in metabolites of the TCA cycle, a reduced dependency on glycolysis, and alterations in lipid metabolism. Additionally, D-2-HG-mediated inhibition of the prolyl hydroxylase domain proteins (EGLN1 and -2) leads to the upregulation of hypoxia-inducible factors (e.g., HIF1α), resulting in a metabolic switch to maintain oxygen homeostasis. D-2-HG also affects collagen maturation via the inhibition of proline and lysine hydroxylases (P4HA1-3 and PLOD1-3), leading to an impaired extracellular matrix structure. Thus, IDH mutations have a wide variety of downstream biological effects; therefore, these mutations are considered as the drivers in multiple tumour types.  [7,44,45]. The family of Jumonji enzymes is also involved in the regulation of the Mechanistic Target Of Rapamycin Kinase (mTOR) signalling pathway, as well as DNA repair via the homologous recombination pathway. Moreover, IDH MUT enzymes have a reduced ability to produce NADPH and consume high levels of NADPH to produce D-2-HG, resulting in severely reduced overall NADPH levels. This deficiency does not only cause metabolic stress but will also lead to an increase in reactive oxygen species (ROS), making IDH MUT tumours more vulnerable to DNA damage. Besides the induction of metabolic stress, IDH MUT tumours also undergo metabolic rewiring, including alterations in metabolites of the TCA cycle, a reduced dependency on glycolysis, and alterations in lipid metabolism. Additionally, D-2-HG-mediated inhibition of the prolyl hydroxylase domain proteins (EGLN1 and -2) leads to the upregulation of hypoxia-inducible factors (e.g., HIF1α), resulting in a metabolic switch to maintain oxygen homeostasis. D-2-HG also affects collagen maturation via the inhibition of proline and lysine hydroxylases (P4HA1-3 and PLOD1-3), leading to an impaired extracellular matrix structure. Thus, IDH mutations have a wide variety of downstream biological effects; therefore, these mutations are considered as the drivers in multiple tumour types.

Inhibition of the IDH MUT Protein
To counteract the oncogenic activity of the IDH mutations, several inhibitors targeting either IDH1 p.R132 variants (e.g., ivosidenib) or IDH2 p.R140 variants (e.g., enasidenib) have been developed over the past couple of years [46]. In vitro studies and clinical trials show that AML patients could benefit from IDH MUT protein inhibitors [26,32], although some patients develop resistance against these inhibitors over time. This acquired resistance is multi-factorial and can be caused by second-site mutations in IDH MUT genes to prevent the binding of IDH MUT protein inhibitors, IDH MUT isoform switching to circumvent the effect of IDH MUT protein inhibitors, or novel acquired mutations in genes encoding for receptor tyrosine kinases (RTKs) [33][34][35]. Direct inhibition of IDH MUT proteins seems less promising for other tumour types that frequently harbour an IDH mutation (Table 1) [27,28,36,37]. Especially in chondrosarcoma, the effect of IDH MUT protein inhibitors in in vitro assays seems controversial. While several studies have shown that IDH1 MUT protein inhibition does not affect the tumourigenic properties of chondrosarcoma cell lines [27,29], other groups have shown that IDH1 MUT protein inhibition causes a decreased proliferation rate in chondrosarcoma cell lines at higher doses or with a different compound [30,31]. Recent results from a phase I clinical trial with the IDH1 MUT inhibitor ivosidenib showed that prolonged disease control (i.e., progression-free survival of~6 months) could be achieved in a subset of patients with advanced chondrosarcoma, predominantly in patients with a minimal number of co-occurring mutations [38]. Together, these results suggest that a subset of chondrosarcomas might have become independent of their IDH mutation over time and that the underlying biological changes either have become static or are driven by other mutations that were acquired later during tumour development.

Synthetic Lethal Interactions with the IDH Mutation
As IDH MUT protein inhibitors showed limited efficacy in in vitro assays and clinical trials or acquired resistance was observed (Table 1), a large number of in vitro studies were performed to determine whether directly targeting the downstream biological effects of IDH mutations would be more promising (Table 2). Indeed, multiple synthetic lethal interactions with the IDH mutation were reported for AML and glioma, including radiotherapy, chemotherapy, and agents that target poly(ADP-ribose) polymerase (PARP), B-cell lymphoma 2 (Bcl-2) family members, Bromodomain and Extra-Terminal Motif (BET) proteins, DNA methyltransferases (DNMTs), mTOR, Nicotinamide Phosphoribosyltransferase (NAMPT), and glutaminase [27,28,[47][48][49][50][51][52][53][54][55][56][57][58][59][60]. However, chondrosarcoma cell lines are variably sensitive to a selection of these therapies, but the effect seems irrespective of the IDH mutation status, as IDH WT chondrosarcoma cell lines show similar treatment responses [61][62][63][64][65][66][67]. These contradictory findings on synthetic lethal interactions with the IDH mutation might be ascribed to different factors. First, the cell of origin and the tumour microenvironment (e.g., cartilaginous matrix formation and hypoxia in chondrosarcoma) of the distinct tumour types that frequently harbour an IDH mutation are highly different and could therefore influence the role that IDH mutations play in tumourigenesis. Second, the level of the D-2-HG oncometabolite may also influence the downstream biological effects of IDH mutations. The most common IDH variants in AML and glioma both produce relatively low D-2-HG levels, whilst the most common point mutation in both cholangiocarcinoma and chondrosarcoma produces relatively high levels of the oncometabolite (Table 1) [17][18][19]. It was recently shown that a lower level of DNA hypermethylation was observed for the IDH1 p.R132H variant compared to non-p.R132H variants, irrespective of tumour type [16]. Lastly, the type of in vitro model (endogenous vs. artificially created) might influence whether synthetic lethal interactions with the IDH mutation are present or not. The introduction of an IDH mutation in a glioma model leads to reduced glutamine and glutamate levels, but this change in TCA cycle metabolites is not present when endogenous IDH WT and IDH MUT glioma models are compared [68]. Most synthetic lethal interactions with the IDH mutation were indeed identified in generic cancer cell lines with an introduced IDH MUT (Table 2). AML and glioma cell lines with an endogenous IDH MUT are scarce, but the utilised chondrosarcoma cell lines do harbour endogenous IDH mutations and this difference in model type could explain why synthetic lethal interactions with the IDH mutation are absent in the chondrosarcoma in vitro studies. As IDH mutations occur early during tumourigenesis, especially in chondrosarcoma, artificial models with an introduced IDH mutation may not be representative of the role that IDH mutations normally play in tumourigenesis. These studies also introduced the IDH mutation in generic cancer cell lines that are easy to transfect (e.g., HeLa, HCT116, and U2OS cells), and these cell lines do not represent the tumour types in which IDH mutations frequently occur. Moreover, most studies generated models that overexpressed the IDH MUT protein, whilst the balanced expression of IDH WT and IDH MUT is needed to retain efficient D-2-HG production [69]. Together, these considerations emphasise that the tumour type, the IDH MUT variant, and the type of in vitro model should be taken into account when studying synthetic lethal interactions with the IDH mutation, and that the underlying vulnerabilities may highly differ between tumour types that frequently harbour an IDH mutation.

Putting the IDH Mutation into Context to Define Underlying Vulnerabilities
In addition to these factors, it was recently shown that the (epi)genetic landscape in which IDH MUT and IDH WT are embedded is another important aspect to take into consideration when defining underlying vulnerabilities in tumour types that frequently harbour an IDH mutation. Studies on AML and glioma have shown that the genetic and epigenetic landscape in which IDH WT and IDH MUT function is highly heterogenous and thereby influences the therapy response and patient outcome [70][71][72][73][74][75][76][77][78][79][80][81]. For instance, mutations in TP53 and ATRX are the underlying denominator in defining which IDH WT and IDH MUT gliomas respond to radiotherapy [70]; the overexpression of BCAT1 in IDH WT AML leads to an IDH MUT -like DNA hypermethylation phenotype [71], and additional mutations in DNMT3A cause reduced levels of DNA hypermethylation in IDH MUT AML samples [74]. Furthermore, co-occurring (epi)genetic alterations such as CIMP status [78], 1p19q deletions [80], CDKN2A deletions [78,79], MET amplifications [78], PDGFRA amplifications [79], and TERT mutations [80] influence overall survival in IDH MUT glioma patients. Moreover, IDH MUT AML patients with a co-occurring NPM1 mutation show overall a better response to chemotherapy with or without venetoclax [81]. The influence of co-occurring (epi)genetic alterations may also explain why distinct IDH MUT tumour types differ in therapy sensitivity and underlines the need to use endogenous IDH MUT models, as generic cancer cell lines with an introduced IDH mutation do not represent the (epi)genetic landscape in which IDH mutations naturally exist. Thus, the IDH mutation status does not solely define the underlying vulnerabilities, which is in line with previous findings for chondrosarcoma [61][62][63][64][65][66][67], suggesting that a dichotomy between IDH WT and IDH MUT is too simplistic.
Besides IDH mutations, chondrosarcomas frequently harbour mutations in TP53, CDKN2A/B, COL2A1, YEATS2, NRAS, and TERT [82][83][84][85][86]. However, the rest of the previously observed co-occurring mutations seem to follow a more random pattern and are present in less than 10% of the chondrosarcomas [25,83,84,87], leading to a highly heterogeneous genetic landscape in which IDH WT and IDH MUT function in chondrosarcoma. Furthermore, IDH MUT chondrosarcomas are characterised by a global hypermethylation phenotype that changes with increasing histological grade [44,45], and, based on methylation profiles alone, several chondrosarcoma subgroups could be defined, even within IDH WT and IDH MUT tumours [88]. Moreover, using chondrosarcoma transcriptome and methylome data, it was previously shown that different molecular subtypes (i.e., high mitotic state, 14q32 miRNA cluster loss of expression, and IDH MUT -induced DNA hypermethylation) exist, and that these are associated with patient outcomes [89]. Moreover, (epi)genetic alterations in the TERT gene (i.e., hypermethylation and promotor mutations) affect the survival probability of IDH1 MUT chondrosarcoma patients, whilst this association is absent in IDH WT and IDH2 MUT patients [87]. Together, these findings show that the IDH mutation status does not solely define the treatment response or outcome in chondrosarcoma patients, suggesting that the dichotomy between IDH WT and IDH MUT is also too simplistic for chondrosarcoma.

Conclusions and Future Directions
Although IDH mutations occur frequently in chondrosarcoma, their prognostic value as well as therapeutic potential seem both ambiguous in chondrosarcoma (Table 1). This is in line with the hypothesis that some chondrosarcomas become independent of their IDH mutations over time and that additional mutations take over the driver role in later stages of tumour development. Nevertheless, other tumour types that frequently harbour an IDH mutation do show the prognostic value of the IDH mutation (glioma) and response to IDH MUT protein inhibitors (AML) ( Table 1). Additional mutations in RTKs could contribute to secondary resistance to IDH MUT protein inhibitors [35], which complements the idea that other mutations can take over the driver role of IDH mutations. As chondrosarcomas are usually diagnosed relatively late due to minimal symptoms in the early stages of tumour development, these additional mutations might have already occurred and may hamper the efficacy of IDH MUT protein inhibitors. This is in line with the fact that an increase in progression-free survival after treatment with an IDH MUT protein inhibitor (ivosidenib) was predominantly observed in chondrosarcoma patients with a minimal number of cooccurring mutations [38]. Thus, the role of IDH mutations most likely differs between distinct tumour types as well as stages of tumour development. Future studies should investigate whether (the number of) additional mutations could be a potential predictive biomarker for the response to IDH MUT inhibition in chondrosarcoma and other IDH MUT tumour types.
This discrepancy in the role of IDH mutations is also reflected in the preclinical studies that have investigated underlying vulnerabilities in IDH MUT tumour types ( Table 2). The contradicting results on synthetic lethal interactions with the IDH mutation between tumour types might be ascribed to a difference in cell of origin, co-occurring (epi)genetic alterations, D-2-HG levels, or the type of in vitro model. The latter could introduce variance between IDH MUT tumour types that does not naturally exist, as these artificially created IDH MUT most likely do not reflect the early onset of IDH mutations or the (epi)genetic landscape in which these mutations are normally embedded. The endogenous IDH WT and IDH MUT chondrosarcoma cell lines harbour additional genetic alterations (unpublished data), including CDKN2A loss, and were derived from patients with high-grade (II and III) and differentiated tumours, meaning that these cell lines are representative models for the role of IDH mutations in more advanced tumour stages. This might explain why these cell lines showed a limited response to IDH MUT inhibitors and showed a variable response to treatments that were identified in other in vitro studies that utilised artificially created models. Future preclinical studies should therefore exercise caution regarding the use of artificially created IDH MUT models and should thoroughly characterise the (epi)genetic landscape in which the IDH mutation was introduced, as well as confirming their findings in models that harbour an endogenous IDH mutation.
Nevertheless, these artificially created IDH MUT models could provide valuable insight into the influence of the (epi)genetic landscape on the downstream biological effects of IDH mutations and thus on underlying therapeutic vulnerabilities. Future studies should expand on the identified subgroups and define novel (epi)genetic aberrations that distinguish subgroups within IDH WT and IDH MUT chondrosarcomas, followed by the identification of tailored targeted therapeutic strategies towards these subgroups. This will improve not only the identification of effective treatment modalities but also the design of clinical trials for high-grade chondrosarcoma patients (e.g., umbrella trial design) and the inclusion of patients in basket trials.

Conflicts of Interest:
The authors declare no conflict of interest.