Current development and future perspective of IDH1 inhibitors in cholangiocarcinoma

Biliary tract cancer (BTC) represents a major public health problem due to its increasing rates of incidence and mortality, especially the intrahepatic cholangiocarcinoma (IHCCA) subtype. First line palliative systemic treatment with cisplatin and gemcitabine has been the unique level IA evidence option until last few years when a deeper understanding of its molecular landscape has unveiled CCA as a very rich targetable disease. This has revolutionised the patient's scenario and has brought new targeted therapies guided by molecular aberrations. Isocitrate dehydrogenase (IDH)1 mutations are the most prevalent targetable alteration in CCA (13% of IHCCA). Ivosidenib has been very recently approved by FDA for IDH1 mutated CCA patients based on a randomised clinical trial (ClarIDHy).


| INTRODUC TI ON
Biliary tract cancer represents 3% of the gastrointestinal tract cancers. 1 Although it is considered a rare disease in western countries (incidence of 0.3-3.5 cases/100.000), its high morbidity and mortality and trending ascend incidence (specially the intrahepatic CCA [IHCCA] subtype possibly related to the increase of obesity and metabolic syndrome) makes of it a major public health problem. 2 BTC is a very heterogeneous disease regarding multiple aspects such as anatomic location, aetiology, natural history, demography, clinical presentation or surgical management. Only around 30% of patients are resectable at diagnosis (the unique potential curative option). The high recurrence rate after surgery even with adjuvant therapy (around 50%) and the frequent advanced stage already at initial diagnosis (70%) make usually palliative systemic treatment the only potentially active option. 3 Chemotherapy based on cisplatin and gemcitabine is still the standard of care in the first line setting and has represented the only 1A level of evidence option for a decade achieving a median overall survival of 11, 7 months in the ABC-02 landmark trial. 4 Last few years have brought new systemic treatment options both in first and second line settings such as new chemotherapy combinations (FOLFOX, 5 Xeliri, 6 gemcitabine-nabpaclitaxel-cisplatine 7 ) and targeted therapies non-biomarker-guided (regorafenib) 8 ( Figure 1).
However, it has been the deeper molecular understanding which has opened a new era for this orphan disease. As many as 40% of the IHCCA harbour molecular alterations which can be triggered. 9 Neurotrophic tyrosine Receptor kinase (NTRK) rearrangement and microsatellite instability-high (MSI-H) are agnostic tumour molecular biomarkers for the treatment with NTRK inhibitors (entrectinib or larotrectinib) 9 and immune checkpoint inhibitors 10 respectively.

Key Points
• BTC, especially IHCCA, has been shown to be a very targetable rich disease.
• IDH1 mutation, previously described in other malignancies, is the most prevalent clinically targetable alteration in IHCCA patients.
• Pathogenic mechanism of IDH mutation is still not well understood. 2HG is the oncometabolite product of the IDH mutation and competes with alpha-ketoglutarate (aKG) causing epigenetic (blocks cellular differentiation) and metabolic dysregulation.
• Ivosidenib, a first-in-class inhibitor of mIDH1, has been approved by FDA in mIDH1 CCA as a biomarker-guided targeted therapy, being the first one with an evidencebased on a randomised clinical trial.
• Primary and secondary resistance mechanisms to IDH inhibitors are now being described.
• New second class more potent IDH inhibitors and novel therapeutically approach looking for potential vulnerabilities related directed or undirected to the IDH mutation, such as targeting DDR or immunotherapy, are in clinical development.
Nonetheless, ClarIDHy has been the first positive randomised clinical trial for a molecular guided-targeted therapy, showing a PFS benefit for ivosidenib (a first-in-class oral IDH inhibitor) vs placebo in the second/ third line treatment of CCA patients with IDH1 mutation. 15 This has led to the very recent approval by FDA. The lower response rate (3%-5%) and frequent primary or secondary resistance events to IDH1 inhibitors merits to keep looking for new treatment options for this population.
In this article, we review the biological basis of the IDH pathway and the oncogenic effects of its mutation in cancer, specifically in CCA. We also review the clinical implications of the IDH1 mutation in terms of diagnosis, prognosis and the current treatment options in CCA. Finally, we explore the potential future treatment options still in progress, attending monitoring of response and resistance mechanism for the IDH mutated CCA patients.

| IDH physiological role
Isocitrate dehydrogenase (IDH) is an essential metabolic enzyme for cellular respiration that participates in the Krebs cycle. IDH has three isoforms called IDH1, IDH2 and IDH3 (the last one has not been described as relevant in cancer). IDH1 is located in the cytosol and the peroxisomes while IDH2 and IDH3 are located in the mitochondria.
The catalytic function of the three enzyme is the conversion of isocitrate into alpha-ketoglutarate (aKG) by an oxidative decarboxylation reaction. IDH1 and IDH2 are Nicotinamide adenosine dinucleotide phosphate (NADP) dependent enzymes and use NADP+ in the reaction that turns into NADPH and CO2. IDH3 is NAD dependent and use NADP+ in the reaction that turns into NADP and CO2 16 ( Figure 2). IDH1 and IDH2 are homodimers while IDH3 is a heterotetradimere. Focusing on IDH1, each unit is composed of 414 amino acids forming a three regions structure (large, small and claps regions). 17 The conformation can adopt two possible and reversible states; open (deactivated form) and closed (activated form). Transition between these two states depends on aKG concentration and on the NADP+/NADPH ratio.
By the mentioned enzymatic reaction, IDH is implicated in multiple cellular processes not only metabolic but also epigenetic.
TET family (5-methyl cytosine hydroxylase) and KDMs are enzymes implicated in DNA and Histone demethylation respectively. This is a crucial step in cell differentiation since gen transcriptional regulation processes will or not allow a cell to express certain genes depending on the context. For instance, maturation of hepatic progenitor cells (cellular differentiation) requires of the expression of HNF4a.
F I G U R E 2 IDH pathway. Wild type and mutant type DEPTOR, the endogenous inhibitory regulator of mTOR, is stabilised by KDM4A (by inhibiting its polyubiquitylation). By this action, DEPTOR decreases mTORC1 and mTORC2 activation (avoiding signals related to cell growth, cell survival, angiogenesis and metabolism).
Regulation of hypoxia inducible factor 1a (HIF1a) pathway depends on PHD and FIH by the hydroxylation of proline and asparagine residues respectively. PHD mediates degradation of HIF1a (by its ubiquitation) while FIH interferes with its catalytic function by avoiding its association with transcription factors such as CBP (CREB binding protein) and p300. This finally controls the hypoxia response. 18 Collagen maturation also depends on prolyl and lysine hydroxylases such as P4HA (prolyl -4-hidroxilase) and PLOD (procollagen lysine 2 oxoglutarate 5 dioxygenase) which collaborate in its correct folding.
ALKBH and also KDM4B belongs to the DNA damage response (DDR) pathway participating in homologous recombination (HR). In such a way aKG also contributes to the genomic stability maintenance. 19 Importantly, IDH1 is the main source of non-mitochondrial NADPH, the other physiological catalytic product. This metabolite participates in different functions. REDOX reactions (such as reduction of glutathione and activation of catalase and citP450) require NADPH to attenuate oxidation damage in cells. 20 Also, NADPH, as an electron donor, participates in the synthesis of many molecules such lipids (steroids, phospholipids, triacylglycerol), amino acids (proline and glutamate) and deoxyribonucleotides.

| IDH mutations and cancer
IDH1 is an oncogene located in 2q34 while IDH2 is in 15q26.1.
Hotspot heterozygous missense mutations (no truncation nor frameshift) of IDH1 were first described in low grade glioma (LGG) and secondary gliobastoma followed by acute myeloid leukaemia (AML). Soon after the IDH1 mutation was related to many other solid tumours.
LGG and secondary glioblastoma (23%), 21 CCA (14%) 22 and AML (7%), 23 are the three most frequent mIDH1 tumours. IDH1 mutations are more frequent than IDH2 and both mutually exclusive. IDH1 mutations affect substitution of arginine residue in 132 position (R132) where the enzyme binds isocitrate. 24 Hotspot mutations in IDH2, out of the scope, affect R172 and R140. 21 These mutations occur mainly in heterozygosis so homodimers turn into heterodimers. 25 Although the exact pathogenesis of the IDH1 mutation in cancer is still a matter of debate, it has been clearly demonstrated that mutations confer a gain of function (neomorphic activity) so that the new heterodimer (R132/WT) has more affinity for NADPH than for isocitrate. mIDH acquires a constant closed (activated) conformation leading to a change in the original reaction with the final generation of an oncometabolite called 2hydroxyglutarate (2HG) (now converting NADP+ to NADPH). 26 Levels of 2HG in these tumours reach 50-100 fold compared to the wild type. Potency for producing 2HG depends on the allelic substitution. 27 Maintenance of the heterozygous state has been reported necessary to maintain closed and activated conformation. 16 Pathogenic consequences of IDH1 mutation are leaded by the increase of 2HG and the reduction of aKG and NADPH ( Figure 2). 2HG has two enantiomers, D2HG and L2HG, being the first one more potent as an epigenetic deregulator. It competes with aKG for the already mentioned aKG dioxygenases, due to its structural similarity (oxidation of C-2 position is the only difference) but with higher affinity. 28 Regarding TET and KDMs, due to the DNA / histone demethylation inhibition, mIDH1 generates a hypermethylator phenotype which causes a gene expression dysregulation (epigenetic "rewiring"). DNA hypermethylation will occur mainly in the CpG islands (low levels of 5-OH-metilcytosine) 29 while histone hypermethylation will generate an increase of methylation markers such as H3K27me3, H3K9me3 and H3K4me3. H3K27me3 and H3K9me3 are transcription repressors and this will affect genes of differentiation such as the previously mentioned HNF4a for hepatoblasts, blocking them in the progenitor state and avoiding the cellular differentiation. 29,30 This probably creates a pathological self-renewal of stem-like progenitor cells (unchecked cell proliferation) that is more permissive to malignant transformation. It has been reported in glioma that IDH1 mutations are related to a colder TME with less CD8 infiltration because of down-regulation of STAT1. 31 Moreover, genes related to the immune response, such as NKG2D, are also silenced in IDH mutant glioma cells, generating an immunosuppressive environment. 32 Since KDM4A stabilises DEPTOR, its inhibition by 2HG increases the level of mTOR by a PTEN-independent mechanism. 33 Also, alterations in the maturation of collagen such as defects in the glycosylation of the OH-lysine of collagen 4 (due to impairment of the already mentioned P4HA and PLOD activity) generate instability and more solubility. This means fragility of the basal membrane of blood vessel facilitating epithelial invasion. 34 Moreover, defects in endostatatine (an antiangiogenic protein that is produced from the collagen XIII and that downregulates HIF1a) generate neoangiogenesis as a consequence. 28 The negative interaction with PHD and HIF leads to an increase of the HIF1a signalling which also activates angiogenesis. 16 Low levels of NADPH, the other consequence of the mutation in IDH, affect the REDOX reactions leading to a higher sensitivity to oxidative stress and to DNA damage agents. Also, NADPH deficiency impairs lipogenesis and the synthesis of amino acids. It has been reported in mIDH1 glioma that decrease in NADPH mitochondrial pool decreases sensitivity to apoptoses 35 and also induces the contribution of glutamine to lipogenesis under conditions of hypoxia. 36 Interestingly, low basal levels of NAD+ in IDH mutated cells, consequence of a decreased expression of the Nicotinate phosphoribosiltransferase (NAPRT1) (enzyme of the NAD+ salvage pathway), disrupts the DDR pathway since PARP utilise NAD+ to generate poly-ADP-ribose (PAR) chains. 37 Also, increase level of H3K9me3 downregulates ATM (damage sensor of DNA). 38 This mechanism is implicated also in genome instability which confers in glioma more sensibility to radiation, alquilating agents and PARP inhibitors. 39 Moreover, D2HG also disrupts enzymes related to the reparation of DNA damage such as previous mentioned ALKBH and KDM4B leading a genomic instability and mutagenesis 39 - The reported biological consequences of the IDH mutations until now do not meet exactly between different tumours suggesting that each cell and its specific genetic context is relevant to it. For instance, the level of the classic methylated phenotype described above is higher in glioma than other mIDH tumours (19% of CpG islands methylated vs 2%-4% respectively). 40

| IDH1 MUTATI ON S IN CHOL ANG IOC ARCINOMA
Improvement in molecular technology has revealed that BTC is a very heterogeneous disease not only at clinical level which classically depends on its anatomic origin but also at a molecular level.
In the last few years several papers have described many different molecular alterations such as mutations, fusions or amplifications between other. Importantly, some of them such as FGFR2, IDH1, HER2, MSI or BRAF are potentially targetable.
IDH mutations in IHCCA were first described in 2012 by Borger.
Most of the IDH mutations in BTC occur in the intrahepatic cholangiocarcinoma subtype which is originated above second degree biliary radical. 22 The incidence of IDH1 and IDH2 mutations in IHCCA is about 14% and 4% respectively being both mutually exclusive 41 ( Heterogeneity of the cohorts and samples issues may explain different results between authors. The majority of knowledge about the cancer pathogenesis of IDH mutations comes from other tumours such as glioma and LMA as we have mentioned before. The overproduction of 2HG both in tissue and blood has also been described in CCA. 49 In a molecular clustering reported by the International Cancer Genome Consortium (ICGC) based on the multiplatform analyses of 489 CCA, IDH mutated tumours were allocated mostly to cluster 4. This cluster showed a hypermethylator phenotype.
Interestingly, cluster 1 also showed hypermethylation but in this case this was directed by down-regulation of TET2 in a similar way to AML. 10   Following diagnosis of IDH mutations, circulating tumour (ct) DNA has been reported such an optimal tool for diagnosis with a very high correlation between tissue and plasma was 84%-92%. [55][56][57] Elevation of 2HG was shown a valuable sensitive and specific marker (levels of 2HG over 170 ng/mL showed 83% and 90% respectively) for diagnosis of mIDH and also correlated with a high tumoral burden. 22 Some authors have reported IDH mutation as a truncal step in the carcinogenesis process. However, others have reported that only in 20% of mutated IDH1 CCA patients, this mutation was already present in the primary tumor suggesting that it could represent a subclonal event instead of a truncal one. 58 In a recent report of a retrospective analysis comparing CGP from primary biopsies of IHCCA to the metastatic ones in 1268 samples, although many findings were common, some of them such as IDH1 mutations and FGFR2 fusion were significantly more frequent in the primary site. 59 The authors concluded that this was suggestive of a wrong primary diagnosis in the metastatic samples.

| MED I C AL TRE ATMENT OF mIDH1 CC A PATIENTS
Publication of the results of the ClarIDHy, the first positive randomised trial for a targeted therapy (using the IDH1 inhibitor IVOSIDENIB in CCA patients with an IDH1 mutation), has opened the pathway not only for using this targeted therapy in a selected group of patients (very recently approved by FDA) but also for doing molecular testing in every advanced CCA patient candidate to systemic treatment, especially in intrahepatic CCA.

| IDH inhibitors: Ivosidenib
Direct IDH inhibition is the most exploited treatment option for mIDH CCA patients. The mechanism of action of IDH inhibitors consists of stabilising the mutated enzyme in its opened conformation (inactivated) by binding tight to its allosteric site (hydrophobic interaction and hydrogen bridges) avoiding the transformation to its closed conformation (activated). 39 Blocking the 2HG production, IDH inhibitors may restore appropriate methylation state, releasing again the cellular differentiation process. 23 Most of these drugs are nonspecific for an isoform. AG5198 showed in preclinical models of glioma a reduction in 2HG, 61 in histone methylation markers and in the cellular proliferation but an increase in differentiation markers expression. 62 Issues with its pharmacodynamics leaded to modify the compound resulting in new compound called AG120 or ivosidenib.
AG120 is a first-in-class oral, once daily, oral, potent, selective, reversible, small-molecule inhibitor of the 5 R132 isoforms of IDH1.
It shows a higher polarity, solubility, permeability and stability inside the liver microsomes. Interestingly, AG120 also inhibits the IDH wild type enzyme. AG120 has lower PXR (human pregnane X Receptor) activation and a less ratio of efflux. 63 Induction of differentiation in

AML cells and inhibition of migration /invasion in chondrosarcoma
cells by AG120 was demonstrated in vitro. 64 Based on the results of a phase 1 trial, ivosidenib was approved by FDA in relapse/refractory AML after demonstrating a CR of 25% with median duration of response (mDOR) of 8.2 months. 65 Interestingly, after QTc prolongation (7.8%), the second most fre-  66 In the CCA subgroup patients (N 73), ivosidenib showed a good tolerability without any DLT and a TRAEs G3-4 rate of 5%. Regarding all grade TRAEs, asthenia (25%), nausea (19%) and diarrhoea (12%) were the three most common. In terms of efficacy the ORR was 5% and the DCR 56%. Median progression free survival (mPFS) and median overall survival (mOS) were 3.8 months and 13.8 months respectively.
Interestingly, the rates of PFS at 6 and 12 months were 38.5% and 20.9% respectively. 67 These results conducted to the Ph3, a pivotal international randomised, double-blind, placebo-control, ClarIDHy trial. 15 In this landmark trial, 186 CCA patients with IDH1 mutation (centrally confirmed) and refractory to one or two previous lines of treatment, ECOG 0-1, and measurable disease, were 2:1 randomised to ivosidenib (500 mg QD) vs placebo in 28 days' cycles. 15  Moreover, it has a nice tolerability profile. The most common G3-4 TEAEs (placebo vs ivosidenib) were ascites (6.8% vs 9%), anaemia (0% vs 7.2%), blood bilirubin increased (1.7% vs 5.4%). Interestingly, although TEAEs leading to dose reduction or interruption were more common in the ivosidenib arm (0% vs 3% and 18.6% vs 30.1% respectively), TEAS leading to discontinuation were more common in the placebo arm (6,6% vs 8.5%). In a postdoc analysis, Ivosidenib has found to preserve physical, cognitive and emotional functioning and also to improve pain from baseline (QLC-Q30 questionnaire). These results have led to the very recent approval by FDA of ivosidenib for CCA patients with IDH1 mutation in PD to standard chemotherapy ( Figure 3).

| Biomarkers
In parallel with the development of new treatments, the validation of novel biomarkers for predicting and monitoring response as well as for understanding the mechanisms of resistance are in progress too (Table 3).

| Predicting response
Besides its value as a diagnostic tool as previously mentioned, ctDNA has been reported such a tool for predicting response.

| 25
ADEVA to antitumor activity in CCA. As we mentioned before, magnetic resonance spectroscopy for glioma spectroscopy can also show monitor 2HG levels. 53

| Mechanisms of resistance
Resistance mechanisms to IDH1 inhibitors have been first described in AML. In a retrospective molecular profiling for AML patients treated with ivosidenib, 14% developed a 2nd site mutation in IDH1 (S280F the most frequent). Interestingly, 12% developed a first site mutation in IDH2 (R140Q). 69 These mutations avoid the binding of ivosidenib and increase 2HG. 70 Regarding to CCA, in the Ph1 trial, 59% of patients underwent  71 In the study reported by Lapin, it was found a median of 1-5 new alterations by each patient, being TP53 and ARID1 the newest frequent found but no new isoforms of IDH were present. 68 Primary resistance mechanism is not known.

| New IDH Inhibitors
In an attempt to improve results of ivosidenib, there are new IDH inhibitors in development in CCA (Table 4).

| IDH305
This drug inhibits two isoforms of IDH1 (R132H and R132C) and has demonstrated a high penetrance in HEB and a low clearance by the liver microsomes. 72 A Ph1 trial is ongoing that includes mIDH1 CCA patients (NCT02381866).

| LY3410738
LY3410738 is a novel first-in-class inhibitor that achieves a covalent and irreversible IDH inhibition and is considered a second generation IDH inhibitor. It does not only inhibit R132H and R132C with high potency (IC50 4,16 nM and 2,35 nM respectively) but also inhibits resistance mutations, such as IDH1m new isoforms (like S280F) or new primary IDH2 mutation (R140Q and R172K). 73 LY3410738 is now being tested in a phase 1/2 clinical trial for mIDH solid tumours including CCA, both in monotherapy (refractory patients) and in combination with cisplatin-gemcitabine (first line setting) (NCT04521686).

| FT 2102 (Olutasidenib)
This drug inhibits 2 isoforms of IDH1 (R132H and R132C) by its binding to the isocitrate pocket avoiding the conformational change to the activated state. It has shown in xenografts a good oral bioavailability and high penetrance of HEB achieving a decrease in 2HG. 74 Olutasidenib is being tested in a Ph1/2 clinical trial that includes CCA patients both in monotherapy or in combination with cisplatin-gemcitabine or with nivolumab (NCT 03684811).

| BAY1436032
This pan-mutant IDH1 inhibitor inhibits the 5 R132 isoforms of IDH1 and was selected after screening of more than 30 million compounds. BAY143602 demonstrated activity both in preclinical models of glioma and AML by reducing 2HG and increasing overall survival although without decreasing tumour size. 63 This efficacy was also clinically in a phase 1/2 of AML patients 75 and it is now being tested in a phase 1 trial of solid mIDH1 tumours (NCT02746081).

| AG-881 (Vorasidenib)
This is a first-in-class pan inhibitor of IDH (IDH1 and IDH2). In preclinical models of AML, AG881 is able to induce myeloid differentiation of blasts. 63 This dual inhibition could represent an opportunity to face the mentioned switch of isoform mutation at resistance. A very recent publication of the LGG cohort of a phase 1 (NCT02492737) for mutant IDH1/2 solid tumours (including CCA), vorasidenib was well tolerated (DLT of elevated transaminases occurred at doses ≥100 mg) and showed preliminary antitumor activity (response rate by RANO of 18%). 76 HMPL-306 is another pan inhibitor of IDH and it is also being tested in a phase 1 trial that includes mIDH CCA patients (NCT04762602).

| Other vulnerabilities for IDH mutated CCA
As we have previously mentioned, IDH1 mutant CCA presents other potential vulnerabilities that could be exploited besides IDH1 inhibitors. We summarise some of the strategies in development (Table 4).

| IDH2 inhibitors
Although IDH2 inhibitors has been developed in AML, we already

| Chemotherapy association
Due to the mainly cytostatic mechanism of action of IDH inhibitors and aiming to increase the ORR there is phase 1 trial on going combining ivosidenib with systemic chemotherapy (cisplatin -gemcitabine) in the first line setting in CCA IDH1m patients (NCT04088188).

| Immunotherapy
As we have mentioned, mIDH CCA seems to be an immune cold tumour. In theory, inhibition of mIDH could reverse this immune state and make the tumour more sensible to immune checkpoint inhibitors.
In fact, in a posthoc analyses of the phase 1 trial previously mentioned of ivosidenib, it was shown that in the subgroup of patients with a cytoplasm decrease after treatment there was also upregulation of several immune responses -related genes such as CTLA4, CXCL10 and CD3G. 79 Moreover, in a mouse model of IHCCA with IDH1 mutation treated with ivosidenib it was detected an enhanced immune response to ivosidenib with an increase of recruitment of CD8-positive T-cells to the tumours and induction of immune stimulatory interferon signalling (PD-L1 expression was also upregulated).
In such a way, there is a strong preclinical rationale for combination treatment strategies with ivosidenib. 80 A ph1 trial is evaluating the combination of ivosidenib with a PD1 inhibitor (nivolumab) in IDH1m solid tumours (NCT04056910).

| DNA damage response (DDR)
As we mentioned before, DDR alterations have been described as associated with IDH mutations due to their effect on aKG dioxygenases. 39  In these cases, IDH WT patients but with high ITH were reclassified as IDH-like tumours. Regarding prognosis, it is interesting that mutations in IDH did not impact OS. However, it did the molecular subtype attending ITH; IDH subgroup patients had worse prognosis than the non-IDH subgroup. Furthermore, regarding immune status, the IDH subgroup (and in parallel high ITH tumours) showed a colder tumour microenvironment (TME) with fewer CD8+ and more neutrophils (P < .01). These results suggest that single region sequencing could not be enough to evaluate the molecular profile of a tumour. 85

| Radiotherapy
It has been reported in preclinical models that the REDOX misbalance due to IDH mutation could potentially make these cells more sensible to ionizing agents like radiation. 20 More research in this field for CCA could bring new options.  These results together with financial toxicity, have make some authors argue against ivosidenib. However, comparison is biased because of the still not well-known prognostic value for the IDH patients. Also, toxicity profile looks better for ivosidenib than for chemotherapy in a population that usually needs a good tolerability profile. Moreover, chemotherapy could be still a reasonable option

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.