Pilot Study of High-Dose Pemetrexed in Patients with Progressive Chordoma

The term chordoma encompasses tumors arising from remnants of the embryonic notochord and is an ultrarare disease with approximately 300 cases per year in the United States (1). The clinical behavior of chordoma is variable, from common chordoma subtypes typically progressing slowly to rare dedifferentiated chordomas behaving aggressively with early and widespread metastases. As these tumors arise along the axial spine and skull-base, chordomas almost universally impinge on critical neurovascular structures, which can lead to severe symptoms ranging from dysarthria, diplopia, and dysphagia to pituitary dysfunction, focal weakness, numbness, pain, and bowel/bladder incontinence. Moreover, existing treatments may have high morbidity where even curative treatments can leave patients debilitated. The treatment for primary and locally recurrent chordoma is en bloc surgical resection or maximally safe surgical resection when feasible followed by radiotherapy. For primary tumors, this can lead to long-term disease control. However, locoregional recurrence occurs in up to 50% of cases and approximately 30%–40% of patients will develop metastatic disease (2–4). Median overall survival from the time of diagnosis is 6–7 years, with the 5- and 10-year survival rates ranging from 47%–76% and 42%–71%, respectively (3–6). Once chordomas recur or metastasize, treatments are palliative, and no systemic therapy has been approved.

Chordomas may be subclassified into conventional, chondroid, poorly differentiated, and dedifferentiated, with these categories having implications for overall survival (7, 8). Pathologic diagnosis is aided by protein markers such as brachyury, S100, cytokeratin, epithelial membrane antigen, and integrase interactor 1 (INI1; ref. 2). Prognostic and predictive markers have been identified for recurrence (9–13) and aggressive features such as bone invasion (14). Molecular aberrations in cell-cycle regulation and receptor tyrosine kinase signaling are also being identified that contribute to tumor growth and progression and have led to several clinical trials of agents targeting these alterations (4, 15). However, no agent has emerged as the preferred regimen of choice, leaving a high unmet need for effective treatment.

We previously demonstrated that over 80% of chordomas do not express thymidylate synthase (TS), an important enzyme in folate metabolism (16). Disrupting folate-dependent metabolic processes essential for cell replication has been a therapeutic strategy for patients with cancer over the last several decades. Methotrexate is an antifolate agent extensively used to treat central nervous system (CNS) malignancies, and approval of a second-generation antifolate agent has brought pemetrexed into the treatment repertoire. In addition to inhibiting dihydrofolate reductase, the target of methotrexate, pemetrexed also inhibits TS, glycinamide ribonucleotide formyltransferase, and to a lesser extent other enzymes involved in purine and pyrimidine synthesis (17). Pemetrexed has demonstrated activity in multiple tumor types and prior work in lung cancer has shown that absence of TS expression correlates with sensitivity to pemetrexed (18, 19). Therefore, we hypothesized that chordoma may respond to pemetrexed, and that higher pemetrexed doses than the approved dosage might increase drug penetration to tumor sites. This concept follows the practice of using high doses of methotrexate to treat primary CNS lymphoma. Furthermore, pemetrexed shows relatively mild toxicity (20), reduced risk of renal toxicity over methotrexate, and has a broader target spectrum, which decreases the likelihood of developing resistance (21). Pemetrexed is approved to treat locally advanced or metastatic non-squamous non–small cell lung cancer and malignant mesothelioma at a dose of 500 mg/m² every 3 weeks. Studies have shown pemetrexed can be administered safely at doses of 900 mg/m² for brain metastases (22–24) and CNS lymphoma (25, 26), 1,800 mg/m² for locally advanced/metastatic breast cancer (27), and 1,910 mg/m² for CNS tumors (28). In our clinical practice, we treated 2 patients with heavily pretreated metastatic chordoma negative for TS with high-dose pemetrexed. We observed exceptional responses, including complete responses in the brain and partial responses in body lesions (29).

Development of minimally invasive biomarkers to monitor disease burden and treatment responsiveness in chordoma can overcome several limitations. For example, percutaneous biopsies can increase the risk of seeding tumors along the biopsy track, and radiographic findings can be confounded by surgical procedures, hardware used to stabilize the spine, and tissue reconstructions. Blood molecular biomarkers such as circulating tumor DNA (ctDNA) and cell-free miRNAs (cfmiR) have been used in early detection, evaluation of therapy response and resistance, and detection of tumor recurrence in various cancers (30–32). Mutations in ctDNA have been detected in plasma of patients undergoing surgical resection of spinal chordoma and in longitudinal postoperative plasma of recurrent cases (33). However, the main issue is that mutation frequency is low, and the mutated genes are variable in patients with chordoma. miRs are small, noncoding RNAs that posttranscriptionally regulate gene expression in multiple cancer types (34), including chordoma (35, 36). Furthermore, miRs can be released by tumors into the blood stream, which allow for repetitive monitoring of patients using minimally invasive assays. Therefore, questions remain for whether cfmiRs detected in plasma of patients with chordoma could be used to develop a signature for early detection and potential treatment monitoring.

Taken collectively, we performed a prospective open-label, single-arm, single-institution pilot study to evaluate the antitumor activity and safety of high-dose pemetrexed in patients with progressive chordoma (NCT03955042). The primary endpoint was objective response rate (ORR) according to RECIST v1.1. Secondary endpoints included adverse events, progression-free survival (PFS), tumor molecular profiles, and alterations in tissue and blood-based biomarkers.

This is an investigator-initiated, open-label, single-arm, single-institution pilot clinical trial conducted at Providence Saint John's Health Center. The clinical trial was registered on ClinicalTrials.gov (NCT03955042).

The trial protocol and all amendments received approval by the WIRB-Copernicus Group Institutional Review Board (IRB) (Protocol JWCI-17-0804). The study was conducted in accordance with the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice Guidelines. All patients provided written informed consent before enrollment.

Eligible patients were 18 years or older with a diagnosis of chordoma, a Karnofsky performance status (KPS) of 50% or higher, and had the ability to take folic acid, vitamin B12, and dexamethasone. Patients had adequate organ function consisting of an absolute neutrophil count ≥ 1.5 × 10⁹/L, platelet count 100 × 10⁹/L, hemoglobin ≥ 9.0 g/dL, total bilirubin ≤ 1.5 × institution's upper limit of normal (ULN), alanine aminotransferase (ALT), and aspartate aminotransferase (AST) ≤ 3 × ULN, and serum creatinine ≤ 1.5 × ULN. For inclusion in this study, patients were required to have recovered from acute toxic effects of any prior therapies and could not have received any investigational agent within 28 days of study initiation. Women of child-bearing potential and men with partners of child-bearing potential must have agreed to use adequate contraception while on study. Patients were excluded if they had third space fluid uncontrolled by drainage; a severe or uncontrolled medical disorder that could impair the ability to receive study intervention; an active bacterial infection requiring intravenous antibiotics at time of initiating study treatment, fungal infection, or detectable viral infection (such as known human immunodeficiency virus positivity or with known active hepatitis B or C); or a history or presence of impaired cardiac function.

Pemetrexed 900 mg/m² was administered intravenously on day 1 of each 21-day cycle until disease progression or unacceptable toxicity. Standard pemetrexed supportive medications included folic acid 1,000 mcg orally, once daily, for at least 5 days during the 7-day period preceding the first dose of pemetrexed and continuing until 21 days after the last dose; vitamin B12 1 mg intramuscularly, starting 1 week prior to the first dose of pemetrexed and every three cycles; and dexamethasone 4 mg orally, twice daily the day before, the day of, and the day after pemetrexed administration. Dose modifications were permitted for toxicity.

Safety evaluations were performed prior to treatment and on day 1 of each cycle, and consisted of clinical laboratory assessments, vital signs, physical exam, performance status assessment, and neurologic exam. Adverse events were monitored throughout the trial and graded according to NCI Common Terminology Criteria for Adverse Events (CTCAE), version 4.03.

Patients were assessed by MRI or CT scans obtained before the first dose of pemetrexed treatment, and after every two cycles. Radiographic response and disease progression were assessed according to RECIST version 1.1. Overall response rate was determined by best response assessed by RECIST v1.1, and PFS was defined as the duration of time from start of treatment until objective tumor progression or death.

Formalin-fixed paraffin-embedded (FFPE) tumors were used for molecular profiling as part of medical management. TS IHC was performed by NeoGenomics Laboratories, with 20% or greater of tumor staining considered positive expression. Tumor DNA (genome sequencing), RNA (RNA sequencing), and protein (IHC) profiling were assayed with Caris Molecular Intelligence (Caris Life Science) or Tempus xE (Tempus, Inc.).

Peripheral blood samples (10 mL) from clinical trial participants were collected in lavender K2EDTA collection tubes (BD Vacutainers) before initiation of pemetrexed treatment, prior to each treatment cycle, and at the end of treatment. Plasma was separated from blood cells by differential centrifugation in a swing bucket centrifuge at 1,300 × g for 10 minutes at room temperature. Plasma samples were transferred to 15 mL polypropylene conical tubes and processed through another round of centrifugation at 1,300 × g for 10 minutes at room temperature to ensure any cellular debris was pelleted. Plasma was then aliquoted, barcoded, and cryopreserved at −80°C.

Peripheral blood samples (10 mL) from normal healthy donors (NHD) were collected in Streck tubes (Streck) and immediately processed. All blood samples were centrifuged twice at 1,600 × g for 10 minutes at room temperature to obtain plasma. Plasma samples were aliquoted under sterile conditions, barcoded, and cryopreserved at −80°C as described previously (32).

FFPE tissues samples collected from patients with chordoma prior to study initiation (n = 7) and from additional patients not participating in the trial (n = 14) but banked under IRB protocols #1095449 and #2017000293 were utilized for HTG miR whole-transcriptome assay (WTA). FFPE tissues were cut into 5 μm sections using rotary microtome HM325 (Thermo Fisher Scientific). All FFPE tissue samples were measured and processed as reported previously (32, 37). Pathologic and clinical variables collected for study patients and additional patients are shown in Supplementary Table S1 and Supplementary Table S2, respectively.

Plasma samples collected prior to treatment with pemetrexed (Pre-T) were available from 14 patients. Plasma samples collected after treatment initiation at cycle 2 (n = 13), cycle 3 (n = 12), cycle 5 (n = 12), cycle 7 (n = 10), and cycle 9 (n = 10) were included in the analysis. A list of samples collected per patient is provided in Supplementary Table S3. Plasma samples collected from NHDs (n = 73) and banked (under protocol #JWCI-18-0401 approved by Providence St. Joseph Health IRB, and protocol #MORD-RTPCR-0995 approved by WIRB-Copernicus Group IRB) were utilized as a control group. Plasma samples from trial participants and NHDs, as well as tissue samples (n = 21), were processed and analyzed using HTG EdgeSeq miRNA WTA next-generation sequencing (NGS). Sequencing was performed on the Illumina NextSeq 550 platform following HTG instructions. Each sample was sequenced with a read length of 1 × 50 bps. The raw FASTQ files generated from sequencing were converted from the Illumina BaseSpace BCL to FASTQ software version 2.2.0 and Illumina Local Run Manager Software version 2.0.0. FASTQ files were analyzed with HTG EdgeSeq Parser software version v5.1.724.4793 to generate raw counts for 2,083 miRs per sample.

The study clinical data were collected and managed using REDCap (Research Electronic data Capture), a secure, web-based software application hosted at Providence St. Joseph Health (38). Safety was assessed for all patients who received at least one dose of pemetrexed. All patients who had a tumor response evaluation and assessment of study agent–related toxicity were included in the intention-to-treat population. Descriptive statistics were used for evaluation of baseline patient characteristics and adverse events. The Kaplan–Meier method was used to estimate PFS and median follow-up, using R software version 4.1.3. This is a pilot study and although no formal statistics were used to calculate sample size for this pilot study, a reasonable sample size of 15 would be adequate to direct future trial development given the rarity of tumor type. As previous trials of sunitinib and imatinib, two agents incorporated into National Cancer Center Network Guidelines for chordoma, showed response rates between 0% and 2% (39, 40), observing a 15% response rate would be considered promising. Thus, a sample size of 15 is similar to the enrollment number of the first stage of a Simon two-stage design, with one or more responses warranting further investigation.

The raw count data from the HTG EdgeSeq Parser were processed, analyzed, and normalized using DESeq2 HTG REVEAL software (version 4.0.1; RRID:SCR_000154). The comparisons to determine differentially expressed (DE) miR across the patients were as follows: (i) NHDs versus pretreatment (Pre-T). The P values in all the comparisons were adjusted by Benjamini and Hochberg's approach for controlling the FDR. miRs were considered DE when the adjusted P < 0.05 and log₂ fold change (FC) > |1.2|. Volcano plots were performed using log₂ FC values and −log₁₀ values of P values between NHDs versus Pre-T. All miR counts were normalized using counts per million (CPM) for data visualization that includes: Venn diagrams, unsupervised hierarchical clustering, principal component analysis (PCA), violin plots, and ROC. In longitudinal analysis using specific miRs for monitoring tumor response, the CPM for each miR were included for samples collected at Pre-T and on treatment (On-T) at the different cycles: C2, C3, C5, C7, and C9. The ROC curves were performed using the CPM values of miR-5739 and the AUC values, sensitivity and specificity were calculated using GraphPad Prism 8 software (GraphPad Software Inc.). For PCA, the CPM values were row scaling by using unit variance scaling method. Principal components were calculated using singular value decomposition (SVD) with imputation which performs imputation and SVD iteratively until estimates of missing values converge. For hierarchical unsupervised clustering, CPM values were ln(x + 1)-transformed. Rows were centered and unit variance scaling was done. Both rows and columns were clustered using Euclidean distance and average linkage. All plots were performed using GraphPad Prism 8 (RRID:SCR_002798) or RStudio software (version 4.2.2). All the figures were unified using CorelDraw (version 2019).

The deidentified patient clinical data generated in this study are available upon request from the corresponding author. The human sequence data were generated at Caris Life Sciences and are not publicly available due to patient privacy requirements. Derived data supporting the findings of this study are available from the corresponding author upon request.

Between February 2020 and June 2021, 15 patients with chordoma were enrolled into the study at Providence Saint John's Health Center. Descriptive analysis of baseline patient characteristics is summarized in Table 1 and individual listing is provided in Supplementary Table S1. The median age was 61 (range, 30–88) years and all patients had progressive disease at the time of enrollment. Primary chordoma tumor sites were clival (n = 8; 53%), sacral (n = 3; 20%), and mobile spine (n = 4; 27%), with all tumors having conventional/chondroid histology and all cases positive for brachyury. Five patients had metastatic disease, with target lesions measured for the study in the lungs (n = 4; 27%) and bone (n = 1; 7%); 10 patients had locally advanced disease. All patients had one or more prior surgeries, most patients had received radiation (n = 12; 80%), and one-third of patients had received prior systemic therapy. The median time between initial diagnosis and study enrollment was 29 months (range, 2–191 months). Prior systemic therapies included imatinib (n = 2), nivolumab plus relatlimab (n = 3), nivolumab plus rapamycin (n = 1), nivolumab plus talimogene laherparepvec plus trabectedin (n = 1), pembrolizumab (n = 1), sirolimus (n = 1), brachyury vaccine (n = 1), afatinib (n = 1), palbociclib (n = 1), and cisplatin (n = 1).

Pemetrexed 900 mg/m² was administered by intravenous infusion on day 1 of each 21-day cycle, along with supportive medications of folic acid, vitamin B12, and dexamethasone. The data cutoff was on July 27, 2022, with a median follow-up of 18 months, and the median number of doses of pemetrexed received by patients was 15 (range, 4–31). Patient P00 discontinued study treatment due to psychosocial issues before a response assessment could be made, patient P06 succumbed to COVID-19, 3 patients (P04, P05, P11) continued to receive the same treatment off-study with their local providers and were censored for disease progression, and 10 patients discontinued study treatment due to disease progression.

Adverse events were generally mild at grades 1–2, the most common being fatigue and nausea. Table 2 summarizes the number of patients with treatment-related toxicities by NCI CTCAE grade. Adverse events occurring in 2 or more patients were fatigue, nausea, rash, constipation, diarrhea, mucositis, vomiting, alopecia, pruritus, lower extremity edema, and alanine transaminase increased. Grade 3 treatment-related adverse events were reported in 2 patients (13%) and included creatinine increased (n = 1) and lymphocyte count decreased (n = 1), and one grade 4 lymphocyte count decreased occurred. Dose interruption due to toxicity occurred in 3 patients (20%) and the pemetrexed dose was reduced to 675 mg/m² in 1 patient due to increased creatinine. No other dose reductions, grade 5 adverse events, unexpected toxicities, or dose-limiting toxicities occurred.

Fourteen patients were evaluable for objective response by RECIST v1.1; 1 patient discontinued study treatment for psychosocial reasons prior to establishing a response assessment. Partial response was observed in 2 patients (14%), stable disease was observed in 10 patients (71%), and 2 patients had progressive disease (14%). Notably, over 60% of patients had a decrease in tumor size from baseline as best response, though only 2 met criteria for partial response (Fig. 1A). Median PFS was 10.5 months [95% confidence interval (CI): 9 months–undetermined] (Fig. 1B). The PFS rate at 6 and 12 months was 67% and 40%, respectively, and the 1-year overall survival rate was 93%. Several patients had clear imaging response (Fig. 2A; Supplementary Fig. S1–S3) with clinical improvement in tumor related symptoms (see below). The other patients had stable disease or progressive disease as best response.

A quality-of-life questionnaire was not included in this study; however, notable anecdotal improvements in tumor-related symptoms occurred in some patients on clinical review. Patient P02 was a 43-year-old female with chordoma of the clivus presenting with blurred vision, double vision, photophobia, and complete blindness in her right eye. After three cycles of treatment, she showed objective improvement in vision, as measured by perimetry tests (Fig. 2B) and disease remained stable while On-T. Patient P04 was a 58-year-old female with chordoma of the cervical spine presenting with dysphagia requiring a gastrostomy tube (G-tube) and a baseline KPS of 70%. She was able to have her G-tube removed after six cycles of treatment and her KPS increased to 80% after 10 cycles. Patient P05 had poor appetite, which improved throughout study treatment as evidenced by weight gain from increased eating, and neck pain, and right sciatica-type pain which also improved throughout study treatment. Patient P14 had significant improvement in back pain and a reduced need of narcotic medications for over 6 months On-T before disease progression.

Bulk DNA and RNA sequencing was performed on available FFPE tumor tissues collected before pemetrexed treatment from 11 patients (Table 3). The mutations identified included FANCA (n = 1), RB1 (n = 1), TERT promoter (n = 1), and SETD2 and WRN (n = 1), and no mutations were observed in six specimens. One sample had HRAS and NY-ESO-1 amplification and SMARCB1 underexpression. Surprisingly, the CDKN2A/B locus deletion was not detected in any of the 11 specimens analyzed. Although MTAP deletions were not assessed, seven specimens had MTAP DNA sequencing data and no gene mutations were observed. IHC revealed TS expression was negative in 11 tumors, positive in three tumors, and was not evaluated in one case. PD-1 was expressed in four out of nine tumors and PD-L1 was positive in one out of 12 tumors. Microsatellite instability was stable in all nine tumors tested, mismatch repair was proficient in all 13 tumors tested, tumor mutational burden was intermediate and low in two and nine out of 11 tumors, respectively. The 2 patients experiencing partial response were both negative for TS expression, and there was a trend for patients with tumors expressing TS to be at increased risk of progressing compared with those with negative TS expression (HR, 4.61; 95% CI: 0.98–21.58; P = 0.053).

To explore the potential of cfmiR as diagnostic markers for patients with chordoma, we compared the cfmiR detection in plasma between clinical trial patients prior to starting pemetrexed (n = 14; Supplementary Table S3) and NHDs (n = 73). All plasma samples were analyzed using HTG EdgeSeq miRNA WTA. DE of 359 cfmiRs with an adjusted P value less than 0.05 was observed in plasma samples from patients with chordoma versus NHDs; 233 cfmiRs were upregulated and 126 were downregulated (Fig. 3A). The top 20 DE cfmiRs (Supplementary Table S4) were analyzed using unsupervised clustering (Fig. 3B) and in a PCA (Fig. 3C). Both analyses suggested that the top 20 DE miRs can distinguish patients with chordoma from NHDs. For example, miR-5739 was the most upregulated cfmiR between NHDs and patients with chordoma (Fig. 3D). In the ROC curve, miR-5739 showed an AUC = 0.993 (sensitivity = 100% and specificity = 95.89, P value < 0.0001; Fig. 3E). In addition, we screened for the top 100 most detected miRs, the top 100 most significantly changing cfmiRs, and the top 100 cfmiRs with the highest FC in the plasma of patients with chordoma compared with NHDs (Supplementary Fig. S4a). We found 38 DE cfmiRs in the plasma of patients with chordoma that meet the criteria (Supplementary Table S5). We proposed that these cfmiRs may also have potential as diagnostic markers for early diagnosis of patients with chordoma.

HTG EdgeSeq miRNA WTA was also performed on tumor tissue specimens collected from patients prior to pemetrexed treatment (n = 7) and additional patients with chordoma not participating in the trial (n = 14). Supplementary Table S3 describes the samples collected by trial participant, and Supplementary Table S2 lists clinicopathologic characteristics of the additional patients with chordoma. A total of 713 miRs were detected in chordoma tissue samples (normalized counts greater than 30; Supplementary Fig. S4b). When comparing the overlapped DE miRs in tissues and cfmiRs in plasma, we observed 207 miR/cfmiRs, indicating that the DE cfmiRs found in plasma correlated with miRs detected in chordoma tissue samples.

To explore whether changes in cfmiR over time might indicate disease progression and response to treatment, we evaluated cfmiR in plasma from 14 patients treated with pemetrexed collected before treatment initiation and before drug administration in cycles 2, 3, 5, 7, and 9 (Supplementary Table S3). While some trends were observed in DE cfmiR in patients during the course of treatment, no statistically significant changes correlated to response in this small dataset (Supplementary Fig. S5).

This investigator-initiated pilot study, despite being opened during the COVID-19 pandemic in 2020, enrolled 15 patients with progressive chordoma who received 900 mg/m² pemetrexed every 3 weeks. Fourteen patients were evaluable for response according to RECIST v1.1. The ORR was 14% with 2 patients achieving a partial response, and 71% of patients demonstrated disease stabilization. The median PFS was 10.5 months. The antitumor activity of high-dose pemetrexed in this study is encouraging for a disease with no approved systemic therapies.

Toxicity was generally mild, with the most common adverse events related to pemetrexed being low-grade fatigue, rash, alopecia, and gastrointestinal events (nausea, vomiting, diarrhea, and constipation). Increased creatinine led to dose reduction in 1 patient, though no drug discontinuations due to toxicity occurred. The dosage of 900 mg/m² every 21 days was chosen to increase drug exposure because safety of this and higher doses was observed in studies of brain metastases (22–24), primary CNS lymphoma (25), solid tumors (28), and our own off-label use in patients with cancer (29). Our study confirms that high-dose pemetrexed is well tolerated in chordoma with an acceptable risk-to-benefit ratio.

Molecular profiling of tumors collected prior to treatment with pemetrexed revealed low tumor mutational burden and microsatellite stability, similar to the findings of Kelly and colleagues (16). PD-L1 was positive in one out of 12 tumors (8%) assessed by IHC, and PD-1 was positive in four out of nine tumors (44%), suggesting PD-1/PD-L1 axis involvement in the microenvironment of a subset of chordomas analyzed. Although not statistically significant, there was a trend of positive TS expression associated with an increased chance of disease progression that will need to be evaluated in a larger sample size.

Mutations in FANCA, SETD2, WRN, and TERT promoter were observed in our study along with loss of SMARCB1, suggesting the involvement of defective chromatin remodeling, defective DNA repair, and increased telomerase activity. Tarpey and colleagues presented the landscape of sporadic chordoma by whole-genome sequencing (WGS; n = 11) or whole-exome sequencing (WES; n = 26) to build upon the previously identified loss of CDKN2A and alterations of PI3K signaling genes as genetic drivers of chordoma (41). In addition to somatic duplication of T, which encodes brachyury, the transcription factor of notochordal development, mutations in SWI/SNF (SWItch/Sucrose Non-Fermentable) complex genes ARID1A and PBRM1, the histone methyltransferase SETD2, and the lysosomal trafficking regulator LYST were observed. Bai and colleagues also reported genomic alterations in PBRM1 and deletions of the CDKN2A/2B locus as the most frequent events in their cohort of 80 skull-base chordomas analyzed by WGS (42). All chordomas in our study (n = 11) had no loss of CDKN2A, though a mutation in SETD2 was observed. In contrast, WES/WGS of 11 extensively pretreated chordomas by Gröschel and colleagues identified molecular alterations associated with impaired DNA repair via homologous recombination, raising the potential contribution of prior treatment to shaping tumor molecular profiles (43). Furthermore, it is unclear how the chordoma site (skull-base vs. sacral vs. mobile spine) contributes to genetic drivers.

To our knowledge, our study is the first to demonstrate that cfmiRs can be concordantly detected in plasma and tissues of patients with chordoma, and to identify genome-wide cfmiRs that are DE in the plasma of patients with chordoma compared with NHDs. The advantages are that the HTG miR WTA is an extraction-free assay that relies in NGS and simultaneously quantifies 2,083 miRs in small amount of plasma. These advantages overcome previous limitations in the field and open new areas of research in the role of miRs as a regulatory factor of chordoma tumors. Furthermore, miRs offer new strategies that, unlike the low frequency tumor mutations, are easier to detect and have a dynamic range to be evaluated in liquid biopsy assays.

Limitations of this study include the small sample size, variable clinical presentation and varied prior treatments, as well as the absence of a control arm. The rarity of chordoma has made clinical trial designs challenging, where previous phase II studies of targeted therapy have had variability in primary endpoints, response criteria, sample size, and disease setting. Primary endpoints have included objective response (39, 40, 44–46), objective response at 12 weeks (47) or 24 months (48), PFS (46), and PFS at 6 months (49, 50) or 9 months (51). While objective response by RECIST v1.1 is a valid surrogate of efficacy in single-arm clinical trials, unidimensional measurements may not adequately capture the complex shape of chordoma lesions and other studies have incorporated bidimensional tumor assessments following Choi response criteria (52), and, more recently, exploratory volumetric analyses for improved characterization of tumor response (46, 48). Use of both RECIST and Choi response criteria in the same study highlights the differences in reported ORRs, such as ORR of 0% versus 33% in an 18-patient study of lapatinib (44), 2% versus 22% in a 43-patient study of imatinib plus everolimus (45), and 4% versus 26% in a 30-patient study of apatinib (46) by RECIST versus Choi, respectively. Approximately half of the phase II studies enrolled only patients with chordoma (40, 44–46, 48, 51), whereas the other half were basket trials of sarcoma and rare cancers (39, 47, 49, 50, 53). The smallest sample size of the phase II studies consisted of nine chordomas treated with sunitinib (39), and the largest sample size to date has been 50 response-evaluable chordomas treated with imatinib (40). Finally, eligibility criteria have been variable for including locally advanced, metastatic, and/or unresectable disease, though the rarity of the disease precludes homogeneous study populations. Considerations regarding the representativeness of study participants are provided in Supplementary Table S6.

Limitations of the cfmiR evaluation are the lack of a second cohort of patients with chordoma that allowed us to validate the performance of the cfmiRs found. A second limitation is that the cfmiRs found may have some implication in resistance to the previous treatment that patients with chordoma received before pemetrexed treatment; thus, their proposal as diagnostic cfmiRs needs to be validated in a treatment-naïve chordoma patient cohort. However, these initial results are promising and suggest that cfmiRs may be potential biomarkers for detecting and monitoring tumor recurrence in patients with chordoma, though further validation is needed in a larger cohort of patients.

In conclusion, our data in this pilot trial demonstrate that high-dose pemetrexed exhibits antitumor activity in a limited sample size of progressive chordoma that warrants further evaluation in larger phase II studies as single agent and/or in combination with other agents, including checkpoint inhibitors. Our data also underscore tumor heterogeneity and suggest effective treatments will require combination strategies to address multiple tumor characteristics. We plan to also incorporate standardized quality of life measures in future studies to quantify the total clinical benefit of this treatment.

S. Kesari reports grants from the Chordoma Foundation and grants and non-financial support from Eli Lilly during the conduct of the study. A. Sharma reports grants from the Chordoma Foundation and grants and other support from Eli Lilly during the conduct of the study as well as personal fees from Novocure and Ronin and other support from Chimerix, Incyte, Apollomics, Blue Earth, Spectrum, Oblato, EpicentRx, Pyramid Biosciences, Boehringer Ingelheim, CNS, and Bavarian Nordic outside the submitted work. M. Nguyen reports grants from the Chordoma Foundation and other support from Eli Lilly during the conduct of the study and grants from Novocure, Chimerix, Oblato, Spectrum, EpicentRx, Pyramid Biosciences, Boehringer Ingelheim, Blue Earth, CNS201, Bavarian Nordic, and Incyte outside the submitted work. J. Truong reports non-financial support from Eli Lilly and the Chordoma Foundation during the conduct of the study and non-financial support from Novocure, Chimerix, Oblato, Spectrum, EpicentRx, Pyramid Biosciences, Boehringer Ingelheim, Blue Earth, CNS, Bavarian Nordic, and Incyte outside the submitted work. R. Nersesian reports grants from the Chordoma Foundation and grants and non-financial support from Eli Lilly during the conduct of the study. E. Rahbarlayegh reports grants and non-financial support from the Chordoma Foundation during the conduct of the study. W. Sivakumar reports personal fees from Stryker and other support from Medical Speakers Network and Zeiss during the conduct of the study. D.S.B. Hoon reports a patent for chordoma assessment miRNA utility pending. L. Anker reports grants from the Chordoma Foundation and grants and non-financial support from Eli Lilly during the conduct of the study. A.S. Singh reports personal fees from Deciphera, Aadi Biosciences, and Daiichi-Sankyo and grants from Tracon Pharmaceuticals and Cogent outside the submitted work; and stock and membership on the Board of Directors, Certis Oncology. T.M. Juarez reports grants from the Chordoma Foundation and grants and non-financial support from Eli Lilly during the conduct of the study. No disclosures were reported by the other authors.

S. Kesari: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing. N. Wagle: Resources, investigation, writing–review and editing. J.A. Carrillo: Resources, investigation, writing–review and editing. A. Sharma: Resources, investigation, writing–review and editing. M. Nguyen: Resources, investigation, writing–review and editing. J. Truong: Resources, investigation, writing–review and editing. J.M. Gill: Resources, investigation, project administration, writing–review and editing. R. Nersesian: Data curation, investigation, project administration, writing–review and editing. N. Nomura: Investigation, writing–review and editing. E. Rahbarlayegh: Investigation, writing–review and editing. G. Barkhoudarian: Resources, investigation, writing–review and editing. W. Sivakumar: Resources, investigation, writing–review and editing. D.F. Kelly: Resources, investigation, writing–review and editing. H. Krauss: Resources, investigation, writing–review and editing. M.A. Bustos: Data curation, formal analysis, investigation, writing–review and editing. D.S.B. Hoon: Formal analysis, investigation, writing–review and editing. L. Anker: Resources, investigation, writing–review and editing. A.S. Singh: Resources, investigation, writing–review and editing. K.K. Sankhala: Resources, investigation, writing–review and editing. T.M. Juarez: Conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing.

This investigator-initiated study was supported by research funding from the Chordoma Foundation and by research funding and study drug from Eli Lilly provided to the institution. We thank the patients who participated in this study. We also thank our generous donors who supported this research with gifts through the Saint John's Health Center Foundation, including support from Will and Cary Singleton, Jamie Siminoff, and the Fritz B. Burns Foundation.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.