Targeting the FGF/FGFR axis and its co-alteration allies

Background We analyzed the FGF/FGFR and co-alteration cancer landscape, hypothesizing that combination therapy might be useful in the presence of co-drivers. Materials and methods We describe FGF/FGFR-altered pathways, prognosis, and co-alterations [cBioPortal (N = 7574)] and therapeutic outcomes [University of California San Diego Molecular Tumor Board (MTB) (N = 16)]. Results Patients whose cancers harbored FGF/FGFR alterations (N = 1074) versus those without them (N = 6500) had shorter overall survival (OS) (median: 23.1 versus 26.4 months, P = 0.038) (cBioPortal). Only 6.1% (65/1074 patients) had no pathogenic co-alterations accompanying FGF/FGFR axis abnormalities. The most frequently co-altered pathways/genes involved: TP53 (70%); cell cycle (58%); PI3K (55%); and receptor tyrosine kinases and mitogen-activated protein kinase (MAPK) (65%). Harboring alterations in both FGF/FGFR and in the TP53 pathway or in the cell cycle pathway correlated with shorter OS (versus FGF/FGFR-altered without those co-altered signals) (P = 0.0001 and 0.0065). Four of 16 fibroblast growth factor receptor (FGFR) inhibitor-treated patients presented at MTB attained durable partial responses (PRs) (9, 12, 22+, and 52+ months); an additional two, stable disease (SD) of ≥6 months (13+ and 15 months) [clinical benefit rate (SD ≥ 6 months/PR) = 38%]. Importantly, six patients with cyclin pathway co-alterations received the CDK4/6 inhibitor palbociclib (75 mg p.o. 3 weeks on, 1 week off) and the multikinase FGFR inhibitor lenvatinib (10 mg p.o. daily); three (50%) achieved a PR [9 (ovarian), 12 (biliary), and 52+ months (osteosarcoma)]. Palbociclib and lenvatinib were tolerated well. Conclusions FGF/FGFR alterations portend a poor prognosis and are frequently accompanied by pathogenic co-aberrations. Malignancies harboring co-alterations that activate both cyclin and FGFR pathways can be co-targeted by CDK4/6 and FGFR inhibitors.


INTRODUCTION
Fibroblast growth factor receptors (FGFRs) are highly conserved transmembrane tyrosine kinase receptors, which regulate basic biologic process such as development, differentiation, cell survival, migration, angiogenesis, and carcinogenesis. 1 FGFR forms a family of four tyrosine kinase receptors (FGFR1-4), and one that lacks an intracellular tyrosine kinase domain (FGFR5). 2 In humans, >20 unique ligands are identified, which are known as fibroblast growth factors (FGFs). 3 Ligand binding to FGFR leads to intracellular phosphorylation of receptor kinase domains, a cascade of intracellular signaling, and gene transcription. 4 Downstream signaling is triggered by intracellular receptor substrates FGFR substrate 2 (FRS2) and phospholipase Cg (PLC-g), leading to subsequent up-regulation of mitogen-activated protein kinase (MAPK), phosphoinositide 3-kinase (PI3K), and signal transducer and activator of transcription (STAT) signaling pathways. 5 Aberrant FGFR signaling as a result of gene amplification, point mutations, and gene fusions has been observed in different tumor types; these alterations are promising targets for cancer therapeutics. 6 An aberrantly activated FGF/ FGFR signaling axis is likely to contribute to tumor growth and proliferation, promote neo-angiogenesis, and participate in acquired resistance to anticancer therapies. Using next-generation sequencing (NGS) with a review of nearly 5000 cancer patients, FGFR aberrations were found in 7.1% of malignancies. 7 There are several non-selective multikinase inhibitors that target FGF/FGFR pathways. Food and Drug Administration (FDA)-approved agents include: ponatinib, regorafenib, pazopanib, and lenvatinib. FGFR alteration status is not required to use those medications. Recently, selective FGFR inhibitors have been investigated specifically for FGFR-altered solid tumors. 8 To date, the FDA has approved erdafitinib for urothelial cancer with FGFR2 or FGFR3 alterations, as well as pemigatinib and infigratinib for cholangiocarcinoma with FGFR2 fusions or other rearrangements. [9][10][11] Although multiple studies have found that the aberrant FGFR signaling pathway is an attractive therapeutic target, FGFR inhibitor-based therapies do not always benefit patients, even if one selects for FGFR-altered cancers. 6 The potential reasons for the variable efficacy of FGFR targeting may be related to multiple factors, but we hypothesized that, in some patients, concomitant oncogenic alterations that appear along with FGFR abnormalities could be associated with primary and/or secondary resistance. 7,12,13 Targeting one specific signal in a complex network of genomic drivers may not be sufficient to control cancer progression.
We have previously shown that dual inhibition of MAPK and cell cycle pathways, when both were co-altered, could be effective, even if single-agent targeting was mostly inactive. 14 This has been suggested to be akin to 'whack a mole'. 15 Furthermore, recent tumor agnostic studies demonstrate that the greater the proportion of genomic alterations targeted, the better the outcome, and studies targeting single alterations may not always show salutary effects. [16][17][18][19][20][21] Combination approaches may conceivably overcome the limitation of singleagent FGFR suppression. However, there are limited data reflecting actual clinical practice regarding matched therapy targeting both the FGFR pathway and the co-alterations.
Herein, we provide evidence that FGFR pathway alterations are associated with a poor prognosis. Even so, co-targeting FGF/FGFR axis alterations and concomitant deleterious alterations showed activity in advanced refractory malignancies.

Somatic alteration identification and annotation
Somatic alterations were identified in tumor tissues by hybrid capture-based targeted DNA sequencing using FoundationOne CDx. 24 Somatic alterations in blood-derived cell-free DNA (cfDNA) were detected by Guardant360. 25 Both assays apply NGS of cancer-related genes. The figure of the landscape of somatic mutations was created using the visualized data feature by the OncoPrinter tool in cBioPortal. 26,27 Somatic mutation annotation of the biological and oncogenic effects was extracted from the OncoKB knowledge base. 28

Molecular matching
The MTB made recommendations to optimize molecular matching for each patient, trying to cover as many aberrant genomic alterations by cognate therapies as possible. 16,18,20,21 Physicians, however, ultimately chose the patient's therapy and they could decide to follow MTB recommendations or not.

Statistical analysis
Summary statistics were used to describe patient characteristics. We evaluated progression-free survival (PFS), which was defined as the time between the start of the treatment and disease progression confirmed by imaging or clinical findings. Overall survival (OS) was measured as the time between the onset of therapy until the last follow-up. Patients without progression at the last follow-up date were censored for PFS at that date. Patients alive at last follow-up were censored for OS at that point. PFS and OS data were represented by KaplaneMeier estimation and the survival endpoints were compared using log-rank tests. OS data from cBioPortal represents the time to metastatic/advanced disease to death or censoring as above. Hazard ratio (HR) was computed by log-rank test and Cox regression analysis. All P values 0.05 were considered significant. All statistical analyses were carried out using R version 4.1.1.

Ethics statement
Our study was carried out according to the guidelines of the UCSD Institutional Review Board [Profile Related Evidence Determining Individualized Cancer Therapy (PREDICT), NCT02478931] and I-PREDICT (NCT02534675) and for any investigational interventions for which the patients consented.
Data for our study will be made available by the corresponding author upon reasonable request.   The landscape of FGF/FGFR pathway and co-occurring alternations in the UCSD cohort of FGFR inhibitor-treated patients (N ¼ 16) is shown in Figure 2B and in Tables 1 and 2 and is similar to those derived from the larger cBioPortal cohort with regard to the commonly co-altered genes/pathways accompanying FGF/FGFR axis aberrations. The median number of alterations in the FGF/FGFR genes was 1 (range, 1-4). Alterations in FGF3, FGF4, FGF19, and CCND1 often . Each column denotes an individual patient, and each row displays a gene. Genetic alterations are color coded by the type of alterations. Coalterations in oncogenic pathways were observed in TP53 pathway genes (70%), cell cycleeassociated genes (58%), receptor tyrosine kinases and MAPK pathwayassociated genes (65%), PI3K signaling-associated genes (55%), and other genomic alteration genes (78%). FGF6, FGF23, FRS2, TERC, ZNF703, CHD4, and PRKCL genes that were found in the UCSD cohort were not listed since they were not assessed in the MSK-IMPACT Clinical Sequencing Cohort. Selected genes of relevant pathways were listed. The diagram was prepared in OncoPrinter tool using cBio cancer genomics portal (http://www.cbioportal.org/). (B) Integrated view of cooccurring somatic oncogenic alterations in 16 cases with FGF/FGFR pathway alterations who received FGFR inhibitors (N ¼ 16, UCSD patients). Each column denotes an individual patient, and each row displays a gene. Genetic alterations are color coded by the type of alterations. Co-alterations in oncogenic pathways were observed in TP53 pathway genes (71%), cell cycleeassociated genes (53%), receptor tyrosine kinases and MAPK pathway-associated genes (35%), PI3K signalingassociated genes (41%), and other genomic alteration genes (76%). The diagram was prepared in OncoPrinter tool using cBio cancer genomics portal (http:// www.cbioportal.org/). MSK-IMPACT, Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets; UCSD, University of California San Diego. co-occurred, probably because they reside on the same amplicon on 11q13. Alterations in MDM2 and FRS2 were mostly concomitant, probably because they reside on the same amplicon on 12q15. Among UCSD patients with FGF/ FGFR pathway alterations, a wide variety of gene coalterations were seen ( Figure 2B). The most frequently co-altered pathways/genes were: TP53 [69% (11/16) Figure 3A) or in the cell cycle pathway (along with the FGF/FGFR genes) (HR: 0.74, 95% CI 0.59-0.92, P ¼ 0.0065; Figure 3B) had significantly shorter OS when compared to patients with FGF/FGFR-altered cancers without those co-altered anomalies. However, patients who had both FGF/FGFR abnormalities and co-alterations in receptor tyrosine kinases or MAPK or PI3K pathway showed no significant difference in OS when compared to patients with cancers harboring FGF/FGFR pathway alterations without those co-altered pathways ( Figure 3C and D).

Characteristics of patients with FGF/FGFR-altered cancers who received FGFR inhibitor therapy
Sixteen patients with FGF/FGFR pathway alterations received FGFR inhibitors (Supplementary Figure S2, PD/SD <6 months PR SD t6 months PI3K pathway, n = 6 (38%)  Table 2). Five patients received monotherapy with an FGFR inhibitor; two achieved SD 6 months/PRdboth with single alterations on NGS [FGFR2 fusion (biliary tract cancer treated with infigratinib) and FGFR K656E mutation (glioneuronal tumor treated with lenvatinib)]; the other patients, who did not respond, had pathogenic genomic co-alterations in their cancers.
Four of 11 patients (36%) who received customized combination therapy achieved SD 6 months/PR; all of these patients had tumors harboring pathogenic coalterations in addition to their FGF/FGFR genomic anomalies. These four patients had malignancies as follows ( Table 2 and Supplementary Table S2, available at https:// doi.org/10.1016/j.esmoop.2022.100647): (i) a biliary cancer with an FRS2 amplification and CDKN2A alteration treated with a combination of the FGFR inhibitor lenvatinib and the CDK4/6 inhibitor palbociclib (achieving a PR for w12 months); (ii) a patient with osteosarcoma whose tumor harbored amplifications in FGF23, FGF6, FRS2 as well as CDK4 and CCND2 amplifications [treated with lenvatinib and palbociclib with an ongoing PERCIST 1.0 (PET) response at 52þ months 29 ]; (iii) an ovarian cancer harboring FGFR4 and CDK6 amplification (also treated with lenvatinib and palbociclib with a PR that lasted w9 months); and (iv) a gastrointestinal stromal tumor (status after four prior systemic therapies) with FGFR1 amplification as well as MYC and ERBB2 amplification, and KIT, ARID1A, and NOTCH2 mutations treated with lenvatinib and pembrolizumab (ARID1A alterations may induce sensitivity to immune checkpoint blockade 30 ) (achieving SD ongoing at 13.1þ months).
Importantly, six patients received a combination of drugs that included the CDK4/6 inhibitor palbociclib (administered because of the presence of alterations in the cyclin pathway) and the multi-tyrosine kinase inhibitor (including FGFR inhibitor) lenvatinib. Three of the six patients (50%) achieved a PR (lasting 9, 12, and 52þ months) ( Table 2). The doses used were palbociclib 75 mg p.o. 3 weeks on/1 week off (approved dose ¼ 125 mg p.o. 3 weeks on and 1 week off) and lenvatinib 10 mg p.o. daily (approved dose ¼ 24 mg p.o. daily). In each case, the lower modified dose was the starting dose, and because tolerability was good and efficacy was seen, the doses were not further titrated upwards. The most common side effects in these patients were grade 1-2 rash and fatigue, which were manageable without dose adjustments. There was one patient who experienced grade 3 neutropenia and required dose modification of palbociclib (dose decreased to 75 mg p.o. 1 week on and 1 week off).

DISCUSSION
Herein we report prognostic and predictive observations related to the presence of FGF/FGFR genomic alterations in a group of pan-cancer patients. Our results indicate that FGF/FGFR axis alterations portend a poor prognosis, as reflected by the observation that patients whose malignancies harbor FGF/FGFR genomic alterations have significantly shorter survival than those with malignancies that are wild type for FGF/FGFR. These results are consistent with prior observations of worse prognosis in patients whose tumors harbor FGFR amplification. 31,32 Importantly, 94% of 1074 cancers harbored genomic coalterations in addition to the FGF/FGFR genomic abnormalities. The most common co-altered pathways/genes were in the TP53 axis (TP53 or MDM2 genes) (70% of cancers); cell cycle (58%); PI3K pathway (55%); and receptor tyrosine kinases and MAPK pathway (65%). Additional genomic alterations were observed in 78% of patients (Figure 2A). The presence of FGF/FGFR axis genomic alterations along with TP53/MDM2 or cell cycle pathway alterations correlated with worse OS compared to individuals with FGF/FGFRaltered cancers without those co-alterations. Co-alterations   Figure 2A and B for the list of selected genes associated with the pathway alterations. CI, confidence interval; HR, hazard ratio; NA, not available.

ESMO Open
in receptor tyrosine kinases or MAPK or PI3K pathway did not show prognostic significance. Sixteen patients with diverse malignancies presented at the UCSD face-to-face MTB 20 had FGF/FGFR pathway genomic alterations and were treated with cognate FGFR inhibitors, either alone or together with agents that targeted pathogenic co-alterations. Targeting FGFR signaling alone or together with co-alterations achieved a clinical benefit rate (SD 6 months/PR/CR) of 38% (6/16) and a median PFS of 14 months among those six patients. Most of the tumors [88% (14/16)] had genomic co-alterations along with the FGF/FGFR pathway abnormalities. These coalterations could conceivably be associated with primary resistance to FGFR inhibition. Indeed, it has been recognized that targeting FGF/FGFR pathway alterations is complicated. 13 Some challenges include differences in antitumor activity among different cancer diagnoses, and variable responses depending on the type of FGFR alterations [fusions and mutations have been reported to achieve higher clinical benefit from FGFR-tyrosine kinase inhibitors (TKIs) when compared to amplification]. 8,33 Moreover, multiple different mechanisms for primary and acquired resistance to FGFR-targeted therapies have been revealed. [34][35][36] We hypothesized that resistance to FGFR inhibitors in FGF/ FGFR-altered cancers could be due, in some cases, to genomic complexity with co-alterations that differ from patient to patient. In our cohort, in line with our hypothesis, 14 of our 16 patients had 1 potentially important molecular coalterations in their cancer. A similar phenomenon was observed in the MSK-IMPACT Clinical Sequencing Cohort.The median number of deleterious co-alterations, excluding FGF/ FGFR alterations, in the UCSD cohort was 5 (range, 0-10).
With regard to FGFR inhibitor therapy, altogether 6 out of 16 of our patients derived clinical benefit [SD 6 months (N ¼ 2)/PR (N ¼ 4)]. The two patients with FGFR alterations (a fusion and a mutation) and no co-alterations responded to single-agent FGFR inhibitors. The other four responders received a combination regimen that targeted FGFR and one or more of the co-alterations.
The most common combination utilized included palbociclib (CDK4/6 inhibitor) (given due to cyclin pathway genomic abnormalities) and the potent multi-tyrosine kinase FGFR inhibitor lenvatinib (given to six patients). Three Months from treatment start 18   of the six patients (50%) attained a PR (duration 9, 12, and 52þ months It has been reported that tumors with FGFR fusions appear to respond to FGFR inhibition. 13,37,38 In our cohort, there were two patients with an FGFR fusion. One patient attained a durable PR, and the other patient did not respond; both patients had biliary cancers. There were several potential reasons for different clinical outcomes. First, the patient who achieved a PR had no co-alterations, while the other patient had concomitant molecular alterations, including in the PI3K pathway, a signal transduction pathway downstream of FGFR pathway. 39 Co-occurrence of FGF/FGFR and PI3K pathway alterations could imply cooperation as potential dual oncogenic drivers. [40][41][42] Therefore, an FGFR inhibitor as a monotherapy would not disrupt oncogenic signaling driven by PI3K pathway alterations. Preclinical studies have reported a high synergistic activity between PI3K and FGFR inhibitors. [43][44][45] However, the early phase combination trial was challenging because of difficulty with tolerability with the high rates of treatment interruptions and dose modification. 46 Another factor may be that, while lenvatinib is a potent FGFR inhibitor, infigratinib (which was the drug administered to the patient who responded) has an even lower IC50 for FGFR (Supplementary Table S1, available at https://doi.org/10. 1016/j.esmoop.2022.100647). Finally, the fusion FGFR2-BICC1 in the case of the patient who attained a PR is known to be a typical oncogenic driver, while FGFR2-BICC2 in the other patient may not be as clear-cut. 8 There are several limitations to the current report. Firstly, the therapeutic arm of the study has a small sample size and a variety of histologies (though the latter may also suggest generalizability of results). Secondly, molecular characteristics of tumors could evolve, particularly with therapeutic pressure, and serial molecular follow-up was not available. Future investigations may require molecular sequential analysis, such as monitoring by circulating tumor cell-free DNA analysis. Thirdly, some FGF/FGFR pathway alterations might not be pivotal for cancer cells, and targeting other accompanied signaling pathways might have a greater effect. Finally, most of the FGFR inhibitors used in this study were not selective to the FGF/FGFR pathway. However, despite these limitations, the current research targeting FGF/FGFR signaling in real-world practice provides evidence for the importance of impacting the products of co-occurring alterations, especially in the cyclin pathway.
In conclusion, we demonstrate that FGF/FGFR-altered tumors have a poor prognosis and frequent co-alterations in several important pathways. Co-targeting of cyclin and FGF/ FGFR alterations with the CDK4/6 inhibitor palbociclib and the FGFR inhibitor lenvatinib can be carried out safely, with responses seen in three of six patients, including ongoing benefit in a patient with refractory osteosarcoma who continues to do well at 52þ months. Further prospective evaluation of this strategy is warranted in patients whose tumors harbor FGF/FGFR pathway alterations.

FUNDING
None declared.