Biotech

Exemestane plus everolimus and palbociclib in metastatic breast cancer: clinical response and genomic/transcriptomic determinants of resistance in a phase I/II trial – Nature Communications

Time Stamp: March 18, 2024 8:00 PM
Source Node: 559271

Patients received limited clinical benefit from triplet therapy

Patients participated in an investigator initiated phase Ib/IIa, open-label clinical trial (NCT02871791) evaluating the safety and efficacy of triplet therapy: palbociclib (a CDK4/6 inhibitor) + everolimus (mTOR inhibitor) + exemestane (steroidal aromatase inhibitor). Eligible patients had been diagnosed with HR + /HER2 − MBC and had progressed on a prior CDK4/6i and a prior endocrine therapy (a nonsteroidal aromatase inhibitor).

Phase Ib MTD/RP2D

A total of 9 patients were recruited into the phase Ib portion of the study between September 12, 2016 and March 27, 2017. All study participants were female. At the starting dose of palbociclib (100 mg), 1 out of 3 patients experienced a DLT (grade 3 neutropenia and grade 2 mucositis). Subsequently, 3 additional patients were initiated at 100 mg, and none experienced a DLT. Palbociclib was increased to 125 mg, and all 3 patients had a DLT (grade 3 neutropenia). Thus, 100 mg palbociclib was declared the MTD. The RP2D for phase IIa was 100 mg palbociclib + 5 mg everolimus + 25 mg exemestane²⁰.

PK profile

The 9 patients in the phase Ib portion of the study were included in the PK analysis (Supplementary Data 1). The mean steady state PK parameters for palbociclib, everolimus, and exemestane was found to be consistent with historical data for each drug given as a single agent, and thus, we did not find evidence of significant PK drug interactions when the three drugs are given concurrently.

Phase IIa Patient characteristics

A total of 32 patients were recruited into the phase IIa portion of the study between May 26, 2017 and June 26, 2019. Patient disposition at the time of data cutoff (Jan 3, 2021) is reported in Supplementary Table 1. All study participants were female, with a median age (range) of 55.5 years (36–73). Most of the participants were initially diagnosed with stage I-III breast cancer and had a disease-free interval longer than two years. However, a substantial percentage of patients (n = 10/32, 31.2%) had been diagnosed with de novo metastatic breast cancer. Bone (n = 28/32, 87.5%), liver (n = 26/32, 81.3%) and CNS + lung + liver (n = 27/32, 84.4%) were the most common sites of metastases (Supplementary Table 2).

Approximately one-third of patients had received one line of prior chemotherapy in the metastatic setting (n = 12/32, 37.5%); the remainder had received no prior chemotherapy (n = 20/32, 62.5%). Among patients who had received prior chemotherapy in the metastatic setting, capecitabine was the most commonly used agent (n = 7/32, 21.9%). Almost all patients had received at least one prior line of endocrine therapy (n = 31/32, 96.9%), and over half had received 2 or more prior lines (n = 17/32, 53.1%). All but 2 patients had received palbociclib (n = 30/32, 93.8%) and the remaining 2 patients had received abemaciclib (n = 2/32, 6.3%).

Efficacy

With a median follow-up (interquartile range) of 23.7 months (20.1–31.8 months), 6 patients had stable disease (SD) ≥ 24 weeks (n = 6/32, 18.8%), 18 patients had SD ≥ 12 weeks (n = 18/32, 56.3%), and 8 patients had progressive disease (PD) (n = 8/32, 25.0%) as best response (Fig. 1B, Supplementary Table 3). The clinical benefit rate (CBR, CR + PR + SD ≥ 24 weeks) was 18.8%, which was below the pre-specified efficacy endpoint of CBR ≥ 65%. Median progression-free survival (PFS) was 3.94 months (95% CI: 3.68–9.63) (Fig. 1C) and median overall survival (OS) was 24.7 months (95% CI: 20.6 – not reached) (Fig. 1D). For the patients that derived and did not derive clinical benefit, PFS was 9.63 months (95% CI: 8.44 – not reached) and 3.78 months (95% CI: 2.00 – 4.70), and OS was 22.6 months (95% CI 16.5 – not reached) and 24.7 months (95% CI: 18.80 – not reached), respectively. Median duration of response (DOR) was 5.42 months (95%CI: 3.74–9.62) and disease control rate (DCR, CR + PR + SD ≥ 12 weeks) was 56.3%.

Safety

The most common all-grade adverse events related to study treatment were neutropenia (n = 29/32, 90.6%), oral mucositis (n = 17/32, 53.1%), thrombocytopenia (n = 9/32, 28.1%), and fatigue (n = 8/32, 25.0%). A total of 25 patients (n = 25/32, 78.1%) experienced neutropenia of grade 3 or higher (Supplementary Table 4). Most treatment-related adverse events were likely caused by palbociclib and/or everolimus in this combination regimen (Supplementary Table 4).

25 patients (n = 25/32, 78.1%) had a dose hold of palbociclib due to toxicity, and 16 patients (n = 16/32, 50.0%) had at least one dose reduction. Furthermore, 22 patients (n = 22/32, 68.8%) had a dose hold of everolimus, and 13 (n = 13/32, 40.6%) had at least one dose reduction due to toxicity. Exemestane was held in 2 patients (n = 2/32, 6.3%) due to toxicity, and no participants required a dose reduction of exemestane (Supplementary Table 5). Patient-level clinical characteristics are in Supplementary Data 3.

Whole exome and transcriptome sequencing of baseline tumor and ctDNA revealed potential resistance mechanisms to CDK4/6 inhibitors and endocrine therapy

As part of the secondary objective of the clinical trial to investigate biomarkers of resistance to CDK4/6i through a correlative multi-omics analysis, we generated WES and RNA-seq data from tumor biopsies and ctDNA samples from patients who participated in the phase II portion (Fig. 2A, Supplementary Data 4, Supplementary Data 5). We collected a research tumor biopsy and blood sample at baseline (after progression on the prior CDK4/6 inhibitor but before initiation of triplet therapy) and additional serial blood samples while on the trial. Additionally, when possible, we acquired archival tumor biopsies that preceded the patient’s initial exposure to a CDK4/6i to serve as a CDK4/6i-naive sample. Given the exploratory nature of this study and our limited sample size, significance tests in our analyses were not corrected for multiple comparisons.

**Fig. 2: Genomic landscape of resistance to CDK4/6 inhibitors in clinical trial baseline biopsies.**

For the baseline tumor biopsies (annotated as T1), WES or RNA-seq was successfully performed and passed quality-control for 18 samples each, with 14 samples having both WES and RNA-seq data (Fig. 2A). WES was successfully performed and passed quality-control for 17 ctDNA samples (12 at baseline, annotated as BB1, and 5 while on therapy, annotated as BBn) from 12 patients, with 11 patients having WES from both a baseline tumor and ctDNA sample. For the pre-CDK4/6i tumor biopsies (annotated as A1), WES or RNA-seq was successfully performed and passed quality-control for 6 and 9 samples, respectively, with 5 patients having WES from both a pre-CDK4/6i and a baseline biopsy. The complete dataset consisted of data from 26 patients: germline-matched WES from 24 tumor samples (19 patients) and 17 ctDNA samples (12 patients), and RNA-seq from 27 tumor samples (22 pts) (Fig. 2A).

We first focused on the genomic landscape (WES) of the baseline tumor and blood biopsies (18 T1’s, 12 BB1’s, n = 19 pts with either a T1 or BB1) (Fig. 2A, Supplementary Data 4, Supplementary Data 5). We identified genomic alterations spanning the spectrum of known genes and pathways previously identified to confer resistance to CDK4/6 inhibitors in 58% (n = 11/19) of patients^3,4,5,7,13 and to endocrine therapy in 74% (n = 14/19) of patients^{11,12,16,17,18}. These pathways, genes, and alterations include: PTEN bi-allelic inactivations (n = 1/19, 5% of patients) and AKT1 activating mutations (n = 4/19, 21%) in the PI3K/AKT/mTOR pathway; NF1 bi-allelic inactivations (1/19, 5%) in the MAPK pathway; ERBB2 activating mutations (n = 2/19, 11%), FGFR1 activating mutations (n = 1/19, 5%) or high amplifications (5/19, 26%), and FGFR2 high amplifications (n = 1/19, 5%) in RTKs; RB1 bi-allelic inactivations (n = 2/19, 11%) and AURKA high amplifications (n = 1/19, 5%) in cell cycle genes; activating mutations in ESR1 (n = 9/19, 47%) in the ER pathway. In addition to these genomic alterations, we also observed loss of ER expression (as measured by immunohistochemistry (IHC)) in baseline tumor samples (n = 3/19, 16%), which we refer to as “loss of ER expression” because these patients were previously diagnosed with ER + /HER2- metastatic breast cancer.

In addition to the alterations in these known resistance genes and pathways, we identified three potential resistance mechanisms in genes belonging to these pathways: an activating BRAF^V600E mutation (MAPK) (in patient 23, Fig. 2B), an activating MTOR^T1977R mutation (PI3K/AKT/MTOR) (in patient 22, Fig. 2C), and activating double PIK3CA^E545K,G1007R mutation (PI3K/AKT/mTOR) (in patient 19, Fig. 2D). All of these mutations were clonal in the baseline biopsy in which they were identified. The activating BRAF mutation was not detected in the pre-CDK4/6 inhibitor biopsy, and thus was acquired/enriched following the treatment with the initial CDK4/6 inhibitor (1 year and 7 months on treatment) or the prior aromatase inhibitor (1 year and 9 months on treatment. We did not have a pre-CDK4/6i biopsy for the other two alterations of interest (MTOR^T1977R, double PIK3CA^E545K,G1007R), so we could not verify if they were acquired/enriched.

To further test the potential of these mutations as drivers of resistance, we looked at whether these samples have other known alterations associated with CDK4/6i resistance. For the activating BRAF^V600E mutation (patient 23), the tumor had no other clonal acquired alterations, but did show loss of ER expression, making it difficult to tease out the degree of this mutation’s resistance effect. For the activating MTOR^T1977R mutation (patient 22), the biopsy also had an FGFR1 high amplification, which partly confounds the role of this mutation in resistance. Note that FGFR1 amplification appears to not always drive resistance to CDK4/6i on its own, since it often co-occurs with other resistance-associated alterations⁵. For the double PIK3CA^E545K,G1007R mutation (patient 19), the biopsy had an activating ESR1 mutation, a known resistance mechanism to endocrine therapy, but did not have any known alterations associated with CDK4/6i resistance. None of these biopsies had additional known oncogenic mutations in the PI3K/AKT/mTOR, MAPK, or RTK pathways (patient 22 had an FGFR1 high amplification, but no known oncogenic mutations), known oncogenic mutations or high-grade copy number alterations (CNA) in cell cycle genes associated with CDK4/6i resistance (RB1, CCNE1, or CDK6), or loss-of-function alterations in FAT1. Overall, the alterations co-occurring with the mutations we identified as potential resistance mechanisms are consistent with their proposed roles as drivers of CDK4/6i resistance, even if some of the co-occurring alterations confound their effect (Fig. 2).

Baseline ctDNA identified actionable genomic alterations not found in baseline tumor samples

For 11 patients, we obtained WES of both baseline tumor biopsies and baseline ctDNA. We leveraged this redundancy in our genomic data to verify the consistency between the tumor and ctDNA WES, and to see what additional information we could learn from having WES from both sources. Focusing on genes in pathways associated with resistance to endocrine therapy and CDK4/6i, we found few differences between the genomic alterations in 9 patients (n = 9/11, 82%) (Supplementary Fig. 1). Most of the differences we identified in these 9 patients were consistent with the higher sensitivity expected from tumor biopsies as compared to blood biopsies, particularly for CNAs. For example, the loss-of-function RB1 mutations in patient 10 (subclonal) and 37 (clonal) were identified in the tumors but not in the ctDNA samples. In addition, we found evidence of biallelic inactivation (loss of heterozygosity and a loss-of-function mutation) for multiple tumor suppressors (PTEN, TP53, NF1) in the tumor but not in ctDNA samples, in which we could only identify the loss-of-function mutations. There were a few cases where some alterations were identified in the ctDNA samples but not the tumors in these 9 patients, all of which involved subclonal mutations or CNAs, and which include a subclonal TP53 loss-of-function mutation and FGFR1 amplifications in 3 samples. We also looked for differences between the identified genomic alterations when looking at mutations with a known oncogenic effect in cancer genes and found no additional differences between the samples of these 9 patients.

For the other 2 patients (n = 2/11, 18%), patients 22 and 32, we identified mutually exclusive clonal driver mutations in cancer genes between the ctDNA and the tumor biopsy pair (Fig. 2C). Each sample pair shared truncal clonal mutations, indicative of a co-existence of multiple tumor lineages in the patient’s cancer. For patient 22, we found a truncal GATA3^M400fs loss-of-function mutation with a clonal activating MTOR^T1977R mutation in the ctDNA sample and a clonal activating double ESR1^Y537S,L536P mutation in the tumor sample. For patient 32, we found a truncal PIK3CA^H1047L mutation with a high-clonality activating ERBB2^L755S mutation in the ctDNA sample and a clonal activating ERBB2^L869R mutation in the tumor sample. The MTOR^T1977R and ERBB2^L755S mutations found only in the blood biopsies are each clinically actionable and are associated with response to mTOR inhibitors like everolimus^21,22,23,24 and the pan-HER kinase inhibitor neratinib, respectively^25,26. Based on this and additional evidence, OncoKB classifies MTOR^T1977R as a mutation with compelling biological evidence (OncoKB Level 4) and ERBB2^L755S as a mutation with compelling clinical evidence (OncoKB Level 3a). Thus, we identified clinically actionable clonal mutations in the ctDNA but not in the tumor baseline biopsy of 2 patients that are each likely to be the mechanism of resistance.

Consistent transcriptomic features in genes associated with resistance to CDK4/6 inhibitors and endocrine therapy

The genomic landscape of baseline tumor biopsies in this trial spanned the spectrum of genes and pathways (RTK, MAPK, PI3K/AKT/mTOR, cell cycle, and ER pathways) known to be associated with resistance to CDK4/6 inhibitors and endocrine therapy (Fig. 2). Motivated by this finding, we hypothesized that genomic alterations in these resistance genes and pathways (in particular, known oncogenic mutations and high-grade CNAs) would have a corresponding high level of transcriptional signature activity (for oncogenic mutations and possibly for high-grade CNAs) or gene expression (for genes with high- grade CNAs). We also hypothesized that some of these genomic alterations and transcriptional signatures could correlate with the intrinsic molecular subtype (PAM50) of tumor samples. To test these hypotheses, we leveraged the genomic and transcriptomic data of the baseline tumor biopsies (14 biopsies with both WES and RNA-seq). Because of the modest size of or cohort, we focused on the genes we and others had previously identified to be associated with resistance (those in Fig. 2), transcriptional signatures from the 50 Hallmark gene sets, and 3 RTK transcriptional signatures from our recent work on resistance to endocrine therapy^11,12. These 3 RTK signatures are associated with transcriptional activity of HER2 mutants (HER2 MUT), FGFR (FGFR ACT), or a combination of both signatures (RTK ACT).

In order to quantify whether the expression of a gene or the activity of a transcriptional signature has a high or low value, we needed a cohort to serve as a reference for gene expression. Given that these tumor samples come from patients with HR + /HER2- MBC, we also needed a reference gene expression cohort that is receptor status-balanced in order to accurately assign a molecular subtype (research-grade PAM50)²⁷. For these purposes, we used the Metastatic Breast Cancer Project (MBCProject), which has genomic (WES, n = 379 tumor samples), transcriptomic (RNA-seq, n = 200 tumor samples) and clinical data (including receptor status: 84 HR + /HER2-, 27 HR + /HER2 + , 10 HR-/HER2+, 12 HR-/HER2-), as the reference cohort²⁸. After assigning a molecular subtype to the 14 biopsies in this trial, we found that 2 samples had a Normal PAM50 subtype, which is indicative of a low tumor content, so we excluded these samples from the joint genomic and transcriptomic analysis (resulting in n = 12 biopsies with both WES and RNA-seq and a non-Normal PAM50 subtype).

In agreement with our hypothesis, joint genomic and transcriptomic analysis revealed a high degree of consistency between the presence of known oncogenic mutations and the activity of transcriptional signatures of these pathways (Fig. 3A, Supplementary Fig. 2A). In particular, activity of each of these transcriptomic signatures was a strong classifier for the presence of oncogenic mutations in the respective pathways, as described in more detail below. Oncogenic mutations were also enriched in tumors with high activity in the transcriptomic signatures. These effects were particularly strong for ESR1 activating mutations in the ER pathway and to a lesser degree with activating mutations in the PI3K/AKT/mTOR pathway (AKT1 or PIK3CA) or the RTK/MAPK pathways (ERBB2 or BRAF) (Fig. 3).

**Fig. 3: Consistency between genomic and transcriptomic features in genes associated with CDK4/6 inhibitors and antiestrogen treatment resistance.**

For the ER pathway, activity in the ER pathway Hallmark signatures were a strong classifier for the presence of ESR1 activating mutations (estrogen response early, AUC = 1.00, P = 1.08 × 10⁻³; estrogen response late AUC = 0.86, P = 2.06 × 10⁻², one-sided Mann–Whitney test) (Fig. 3A, top). ESR1 activating mutations (n = 6/12, 50% of biopsies) were enriched in tumors with high activity in the estrogen response early signature (n = 6/12, 50%; P = 2.16 × 10⁻³, two-sided Fisher exact test). This effect was similar but not statistically significant for the estrogen response late signature (n = 6/12, 50%; P = 6.06 × 10⁻², two-sided Fisher exact test).

For the PI3K/AKT/mTOR pathway, combined activity of mTORC1 signaling and PI3K/AKT/mTOR signaling signatures was statistically significant as a classifier for the presence of grouped activating PI3K/AKT/mTOR pathway mutations (only AKT1 and PIK3CA mutations for these samples) (AUC = 0.83, P = 3.25 × 10⁻², one-sided Mann–Whitney test) (Fig. 3A, middle top). None of these comparisons were statistically significant when looking at each of these signatures or activating mutations individually, or when looking at tumors with high activity in these signatures (all comparisons P > 5.00 × 10⁻²), although activity in the mTORC1 signaling signature was borderline statistically significant as a classifier for activating AKT1 mutations (AUC = 0.85, P = 5.00 × 10⁻², one-sided Mann–Whitney test).

For the RTK/MAPK pathway, activity of the RTK ACT or HER2 MUT signatures was each a good classifier for the presence of grouped activating ERBB2 or BRAF mutations (AUC = 0.96, P = 9.09 × 10⁻³, one-sided Mann–Whitney test for each signature) (Fig. 3A, middle bottom). None of these comparisons were statistically significant when taking each of these activating mutations individually and there was no statistically significant enrichment of these mutations when looking at tumors with high activity in RTK/MAPK signatures (all comparisons P > 5.00 × 10⁻²). Other known RTK/MAPK alterations (NF1 biallelic inactivation, FGFR1 activating mutations and high amplifications) were not associated with high activity in the RTK/MAPK signatures and seemed to inversely correlate with these signatures (Fig. 3A, Supplementary Fig. 2). As detailed below, the lack of association with RTK/MAPK signatures in these individual cases could be related to the presence of concurrent activating ESR1 mutations in those tumors (17_T1, 21_T1, 22_T1, 35_T1).

As an additional consistency check, we verified that deactivating TP53 alterations (biallelic inactivation, n = 2/12, 17%; deep deletions, n = 1/12, 8%) were associated with P53 pathway activity (Fig. 3A bottom). Activity in the P53 pathway signature was a statistically significant classifier for the presence of grouped TP53 biallelic inactivation and deep deletions (AUC = 0.93, P = 1.81 × 10⁻², one-sided Mann–Whitney test). Similarly, these TP53 alterations were enriched in samples with low P53 pathway signature activity (P = 4.54 × 10⁻², two-sided Fisher’s exact test).

Given the observed association of activating ERBB2 and BRAF mutations with RTK/MAPK pathway transcriptional signature, we hypothesized this association could be related to the PAM50 subtype of these tumor samples. Consistent with this hypothesis, grouped ERBB2 and BRAF mutations were exclusive to tumors with a HER2-Enriched (HER2-E) PAM50 subtype (n = 4/12, 33%), in which they were enriched (P = 1.82 × 10⁻², two-sided Fisher exact test), and correlation to the HER2-E PAM50 centroid was a good classifier for the presence of these mutations (AUC = 0.93, P = 1.82 × 10⁻², one-sided Mann–Whitney test) (Fig. 3B). In addition, correlation to the HER2-E PAM50 centroid was a good classifier for the absence of activating ESR1 mutations (AUC = 0.92, P = 7.58 × 10⁻³, one-sided Mann–Whitney test). Unlike the HER2- E PAM50 centroid, the RTK ACT or HER2 MUT signatures were not statistically significant classifiers for activating ESR1 mutations (P > 5.00 × 10⁻²). Activating ESR1 mutations were exclusive to samples with a Luminal A (n = 2/12, 17%) or B (n = 6/12, 50%) subtype and were absent in the tumors with ERBB2 and BRAF mutations. Mutual exclusivity between activating mutations in ESR1 and ERBB2 or BRAF is consistent with prior work in HR + /HER2- MBC^5,11,12,16. Only very recently has a similar association been reported between activating mutations in the RTK/MAPK pathway or ESR1 and the PAM50 molecular subtype of a tumor in HR + /HER2- MBC²⁹.

To verify the consistency between high-grade CNAs (high amplifications, deep deletions) and gene expression, we looked at whether tumors with high-grade CNAs had a corresponding high or low level of gene expression. For 9 out 10 high-grade CNAs there was a corresponding high or low level of expression in these genes (Fig. 3C), which included FGFR1 (n = 2/12, 17%), ESR1 (n = 1/12, 8%), ERBB2 (n = 1/12, 8%), IGF1R (n = 1/12, 8%), and AURKA (n = 1/12, 8%), among others.

Conversely, for the majority of cases with an above-average level of expression in these genes, there were no corresponding high-grade CNAs (52 total gene/tumor cases, 9 of which had a high-grade CNA), a result that is consistent with recent work³⁰. These results support the often-used assumption that high-grade CNAs in resistance- associated genes have a strong effect on their expression but highlight how above-average expression levels do not often have an associated high-grade CNA.

In summary, we found that multiple genomic resistance mechanisms in the ER, RTK/MAPK, and PI3K/AKT/mTOR pathways, such as activating mutations in ESR1, ERBB2, or AKT1 and FGFR1 amplifications, have specific transcriptomic features (pathway transcriptional signatures, gene expression levels) associated with them. Specifically, there was a statistically significant association between ESR1 mutations and ER pathway Hallmark signatures, ERBB2/BRAF mutations and the RTK ACT or HER2 MUT signatures, and the presence of ERBB2/BRAF or absence of ESR1 mutations and the HER2-E centroid correlation. These results suggest that these transcriptomic features could be used to identify the pathways driving CDK4/6i and endocrine resistance in a tumor.

Combined genomic and transcriptomic features in baseline biopsies identify likely mechanism of resistance to prior CDK4/6 inhibitor and endocrine therapy treatments in nearly every patient

The genomic and transcriptomic analysis of baseline biopsies identified known and potential features associated with the resistance to CDK4/6 inhibitors and endocrine therapy. These features include activating mutations in ESR1, ERBB2, FGFR1, AKT1, BRAF, and MTOR; double PIK3CA activating mutations; loss-of-functions mutations in NF1, RB1, and PTEN; amplifications in FGFR1, FGFR2, ERBB2, and AURKA; ER loss; and transcriptional features such as the HER2-E subtype, high transcriptional activity of ER signatures, or high RTK activity (high expression of RTKs or high RTK signature activity).

Based on these and additional features, we asked whether these known and plausible resistance mechanisms could explain each baseline tumor’s resistance to prior CDK4/6i and antiestrogen treatments. Additional features we considered include known resistance features such as the Basal PAM50 subtype¹⁰ and plausible resistance features, some of which have preliminary studies that associate them to reduced sensitivity to CDK4/6i and endocrine therapy, such as amplification or high expression of IGF1R or INSR^12,31,32, and low ER activity (low ESR1 expression or low ER pathway signature activity)³.

For 22 out of 23 patients with WES and/or RNA-seq of a baseline biopsy (n = 22/23, 96%), we identified genomic or transcriptomic features that could explain the tumor’s resistance to CDK4/6i or anti-ER treatments (Fig. 4, Supplementary Fig. 3, Supplementary Data 6). Fourteen patients (n = 14/23, 61%) had known resistance mechanisms to both of these treatments (i.e., published evidence of a mechanistic role in resistance to these treatments), 3 patients (n = 3/23, 13%) had a combination of known and plausible mechanisms, and 5 patients (n = 5/23, 22%) had only plausible mechanisms (i.e., preclinical or theoretical evidence suggesting a role in resistance). Given that the resistance mechanisms identified for each patient can be based on clinical, genomic, or transcriptomic features, we looked at how often each mechanism explained the tumor’s treatment resistance. In particular, we focused on the cases where oncogenic mutations in resistance genes, clinical features, or transcriptomic features alone could explain treatment resistance (Fig. 4).

**Fig. 4: Clinical, genomic, and transcriptomic features can explain resistance to CDK4/6 inhibitors and antiestrogen treatment in patient’s tumors.**

The chart encodes whether key features associated with resistance to CDK4/6 inhibitors or anti-estrogen treatment are seen in the trial baseline biopsies of patients with either WES or RNA-seq of a baseline sample (n = 23 patients). For each patient, it includes whether these features could explain if the baseline tumor is resistant to prior CDK4/6i and antiestrogen treatments, and if these are known or plausible resistance mechanisms. For 22 out of 23 patients a known or plausible resistance mechanism was identified. Features included are based on clinical assays (loss of ER measured by IHC), genomic data (known oncogenic mutations and high-grade amplifications in known or plausible resistance genes), and transcriptomic data (Basal subtype, HER2- enriched subtype, high expression or activity in an RTK gene or RTK signature, high/low expression or activity in ESR1 or an ER pathway signature). Features are colored based on the resistance signaling pathway they are associated with. A modified version of this figure that groups these features by gene and signaling pathway, and includes the data types available for each patient is included in Supplementary Fig. 3. The transcriptional signature activity of the Hallmark gene sets and RTK transcriptional signatures for these tumors is shown in Supplementary Fig. 4.

For 13 patients (n = 13/23, 57%), resistance to both treatments could be explained solely by considering oncogenic mutations in known (ESR1, AKT1, PTEN, NF1, FGFR1, ERBB2, RB1) or plausible resistance genes (BRAF, MTOR, double mutations in PIK3CA), or ER loss (pts. 23, 25, and 31). For 5 patients (n = 5/23, 22%; 4 with no genomic data), the resistance mechanisms identified were solely transcriptomic features: Basal PAM50 subtype (along with high RTK signature activity and low ER activity) in patient 29, high RTK activity (IGF1R expression) and ER activity (high estrogen response early signature activity) in patient 36, and high RTK activity in patients 20 (high FGFR1 expression), 33 (high IGF1R expression), and 41 (high activity of the RTK ACT and HER2 MUT signatures). For 4 patients (n = 4/23, 17%), the resistance mechanisms identified included a combination of oncogenic mutations, high-grade CNAs, and transcriptomic features. For example, FGFR1 amplification and RB1 loss-of-function for patient10; an activating ESR1 mutation and high activity of the RTK ACT signature for patient 24; and IGF1R amplification and high IGF1R expression, a GATA3 loss-of-function mutation, and low ER activity (low signature activity of the estrogen response early and late signatures) for patient 26.

In summary, we found that in 22 out of 23 patients, we could identify known or plausible resistance mechanisms to the prior CDK4/6i and antiestrogen treatments by using the clinical, genomic, and transcriptomic features in their baseline tumors. This includes 5 patients in which the only mechanism identified was a transcriptomic feature, illustrating how transcriptomic data can provide complementary information not present in genomic data and provide insights even for cases when no genomic data is available.

Evolutionary analysis revealed convergent and divergent paths to CDK4/6i + anti-ER treatment resistance in tumors with distinct lineages

For 6 patients for whom we had WES from paired pre-treatment and post-resistance tumor samples (5 patients) or concurrent post-treatment tumor and ctDNA samples from distinct tumor lineages (2 patients), we wanted to identify the genomic alterations that were acquired or became enriched following the prior CDK4/6i treatment and that could be driving treatment resistance. To do this, we carried out evolutionary analysis (tumor phylogeny, clonal dynamics) to identify acquired or enriched genomic alterations in the trial baseline sample(s) (which are post-CDK4/6i) as compared to the older, pre- CDK4/6i sample (Figs. 5 and 6). Note that we refer to high-clonality genomic alterations present in the post-CDK4/6i but not detected in the pre-CDK4/6i sample as acquired alterations, even though they might exist in the pre-CDK4/6i sample or prior tumors with very low clonality. To help contextualize the treatments these samples have been exposed to (and, putatively, become resistant to), we also included a clinical case history for these patients (treatment sequence and duration, timing of biopsies and diagnoses) (Supplementary Data 7).

**Fig. 5: Tumor evolutionary analysis and clinical vignettes for patients who derived clinical benefit from palbociclib, everolimus, and exemestane triplet therapy.**

**Fig. 6: Tumor evolutionary analysis and clinical vignettes for patients who did not derive clinical benefit from palbociclib, everolimus, and exemestane triplet therapy.**

For all patients with a pre-CDK4/6i sample, we identified acquired or enriched genomic alterations in known resistance genes. For patient 32, the post-resistance tumor sample had a clonal ERBB2^L869R activating mutation that was also present in the pre-treatment sample, as well as an acquired ERBB2 focal high amplification (copy number above ploidy from 1.9 to 9.7) not found in the pre-treatment sample. We did not detect ERBB2^L869R in the concurrent post-resistance ctDNA sample, and instead identified an acquired high-clonality ERBB2^L755S activating mutation that was not found in the pre-treatment sample (Fig. 5A). For patient 17, the post-resistance sample showed an enrichment in an activating ESR1^D538G mutation, which was subclonal in the pre-CDK4/6i samples but clonal in the post-CDK4/6i sample (Fig. 5B). For patient 23, the post-resistance sample had shared truncal mutations (PIK3CA^E545K, CDH1^A824fs, and others) and an acquired clonal BRAF^V600E activating mutation (Fig. 6A). For patient 21, the post-resistance sample had shared truncal alterations (mutations of unknown significance and a FGFR1 focal high amplification) and an acquired clonal ESR1^D538G and AKT1^E17K activating mutations (Fig. 6B). For patient 26, the post-resistance sample had shared truncal alterations (SNVs of unknown significance and an IGF1R focal high amplification), an acquired subclonal GATA3^-412fs truncating mutation, an acquired clonal ESR1^H524L missense mutation, and an increased amplification in ESR1 (copy number above ploidy from 5.9 to 7.6) that included the region with ESR1^H524L (Fig. 6C).

Of these patients, two had acquired or enriched genomic alterations only in the ER pathway (ESR1^D538G in Pt 17, Fig. 5B; GATA3^-412fs, ESR1^H524L, and ESR1 high amplification in patient 26, Fig. 5C) and no known acquired alterations associated with CDK4/6i resistance, which suggests that the mechanisms of resistance is primarily related to the endocrine partner.

For patients 22 and 32, in their baseline blood and tumor biopsies, we had identified mutually exclusive clonal driver mutations (Fig. 2C), evolutionary analysis confirmed the co-existence of multiple tumor lineages in each of these cancers (Fig. 5A, C).

For patient 32, the identified tumor lineages showed convergent evolution of ERBB2 activation through distinct known activating mutations and the focal high amplification of ERBB2 (Fig. 5A). The baseline tumor sample is dominated by a lineage with ERBB2^L869R and the baseline blood sample by a lineage with ERBB2^L755S, both sharing a truncal PIK3CA^H1047L mutation. The pre-CDK4/6i sample had clonal ERBB2^L869R and PIK3CA^H1047L mutations, and thus, shared the same lineage as the baseline tumor sample. Unlike the baseline samples, the pre-CDK4/6i tumor did not show an amplification of ERBB2, consistent with the original HR + /HER2- MBC diagnosis. The activation of ERBB2 in the baseline tumor is also reflected in its transcriptional features: a HER2-E PAM50 subtype and high activity in the RTK ACT and HER2 MUT signatures. To our knowledge, convergent evolution of ERBB2 activation following CDK4/6i and endocrine therapy treatment has not been previously reported. (Figs. 5, 6)

For patient 22, we identified two tumor lineages with clonal drivers mutations in divergent pathways: a double ESR1^Y537S,L536P mutation (ER pathway) in the tumor sample and an MTOR^T1977R mutation (PI3K/AKT/mTOR pathway) in the ctDNA sample (Fig. 5C). Both samples had a shared truncal GATA3^M400fs mutation and a shared FGFR1 high amplification, which was additionally classified as a focal high amplification for the tumor sample. Each of the baseline samples had known or plausible resistance mechanisms to CDK4/6 inhibitors and endocrine therapy (ESR1^Y537S,L536P and FGFR1 amplification for 22_T1; MTOR^T1977R and FGFR1 amplification for 22_BB1). Although we could not obtain a pre-CDK4/6i sample for this patient, we did verify that both lineages were observed in metastatic biopsies taken after CDK4/6i treatment and before beginning the trial regimen. These biopsies were profiled using targeted DNA sequencing (OncoPanel) and were from distinct metastatic sites than the baseline biopsy (liver and soft tissue vs. bone for the baseline biopsy). The soft tissue biopsy had a double ESR1^Y537S,L536P mutation, an FGFR1 high amplification, and a TP53 homozygous deletion, consistent with what we identified in the baseline tumor biopsy. The liver biopsy had an MTOR^T1977R mutation and an FGFR1 high amplification, consistent with the ctDNA sample.

Overall, tumor evolutionary analysis identified acquired genomic alterations that we attributed to acquired CDK4/6i and endocrine therapy resistance. These acquired genomic alterations were in the RTK (ERBB2 amplification, ERBB2^L755S), MAPK (BRAF^V600E), PI3K/AKT/mTOR (AKT1^E17K), or ER (ESR1^H524L, ESR1^D538G, GATA3^-412fs) pathways. Finally, we identified co-existing tumor lineages with distinct driver mutations in resistance genes in two patients. In one of these patients, there was strong evidence for the convergent evolution of ERBB2 activation following CDK4/6i therapy.

Response to combined palbociclib, everolimus, and exemestane was correlated with activation of the mTOR pathway

We looked for genomic and transcriptomic features in the baseline biopsies (n = 23 patients with either WES or RNA-seq of a baseline biopsy) that correlated with subsequent clinical benefit to the clinical trial of combined palbociclib + everolimus + exemestane (n = 4/23, 17%). Although there were no definite genomic correlates associated with the group that derived clinical benefit, (which was not surprising given the limited sample size), we did identify alterations in the PI3K/AKT/mTOR pathway as a plausible correlate. There were no clear group differences at the gene level, and many of the known mutations in resistance genes were found in both groups, including activating mutations in ESR1, AKT1, and ERBB2, and amplifications in FGFR1 and ERBB2 (Fig. 2). When looking at genomic alterations in pathways, we found that the tumors in the clinical benefit group all had a known oncogenic mutation in the PI3K/AKT/mTOR pathway (Fig. 7A). For patient 22, the ctDNA sample (22_BB1) had a clonal activating MTOR^T1977R mutation, which is consistent with the evidence linking this mutation with response to mTOR inhibitors such as everolimus^21,22,23,24 (Figs. 2B, 5C). For patient 19, the ctDNA sample had a clonal activating double PIK3CA^E545K,G1007R mutation (19_BB1, Fig. 2D). Patients 17 and 32 had clonal activating AKT1^E17K and PIK3CA^H1047L mutations, respectively (17_T1 and 32_T1, Fig. 2B). Given that everolimus targets the PI3K/AKT/mTOR pathway, this suggests the observed association between response and PI3K/AKT/mTOR mutations could be attributed to the everolimus component of combination therapy.

**Fig. 7: Correlation between mTOR pathway activity and response to triplet therapy in baseline tumor samples.**

Given the observed association between PI3K/AKT/mTOR mutations and clinical benefit, we looked at possible reasons why this association was not observed for all the activating AKT1^E17K mutations. Two of the four patients with an AKT1^E17K– mutant tumor had stable disease but discontinued the trial due to toxicity (Fig. 7A); of the remaining two patients, one derived clinical benefit (patient 17, stable disease >24 weeks) while the other patient did not (patient 21, best response was progressive disease). Their main noteworthy genomic/transcriptomic difference is in their HER2 MUT signature activity (high for patient 21 and low for patient 17; both were ESR1-mutant, RTK/MAPK mutant, and had a Luminal B subtype). Thus, we were only able to fully evaluate whether AKT1^E17K-mutant tumors responded to combined palbociclib, everolimus, and exemestane for two cases and found a clinical benefit for one of these cases.

Consistent with the association between PI3K/AKT/mTOR mutations and clinical benefit, transcriptomic analysis identified a correlation between mTORC1 pathway activity and clinical benefit. mTORC1 signaling signature activity was higher in samples with clinical benefit (P = 2.73 × 10⁻², two-sided Welch’s t-test) and was top 2 among all Hallmark signatures (Fig. 7B, C, Supplementary Data 8). A similar result was found when looking at enrichment of high or low activity of Hallmark signatures and clinical benefit, that is, samples with high activity in mTORC1 signaling were enriched in those associated with clinical benefit (P = 2.20 × 10⁻², two-sided Fisher exact test) (Fig. 7A, D, Supplementary Data 8).

In summary, we found early evidence that clinical benefit to palbociclib + everolimus + exemestane is correlated with activation of the mTOR pathway, namely, PI3K/AKT/mTOR mutations and mTORC1 pathway activity. (Fig. 7).

SEO Powered Content & PR Distribution. Get Amplified Today.
PlatoData.Network Vertical Generative Ai. Empower Yourself. Access Here.
PlatoAiStream. Web3 Intelligence. Knowledge Amplified. Access Here.
PlatoESG. Carbon, CleanTech, Energy, Environment, Solar, Waste Management. Access Here.
PlatoHealth. Biotech and Clinical Trials Intelligence. Access Here.
Source: https://www.nature.com/articles/s41467-024-45835-6

The genome sequence of the Violet Carpenter Bee, Xylocopa violacea (Linnaeus, 1785): a hymenopteran species undergoing range expansion – Heredity

September 15, 2024 8:00 pm

Exemestane plus everolimus and palbociclib in metastatic breast cancer: clinical response and genomic/transcriptomic determinants of resistance in a phase I/II trial – Nature Communications

Patients received limited clinical benefit from triplet therapy

Phase Ib MTD/RP2D

PK profile

Phase IIa Patient characteristics

Efficacy

Safety

Whole exome and transcriptome sequencing of baseline tumor and ctDNA revealed potential resistance mechanisms to CDK4/6 inhibitors and endocrine therapy

Baseline ctDNA identified actionable genomic alterations not found in baseline tumor samples

Consistent transcriptomic features in genes associated with resistance to CDK4/6 inhibitors and endocrine therapy

Combined genomic and transcriptomic features in baseline biopsies identify likely mechanism of resistance to prior CDK4/6 inhibitor and endocrine therapy treatments in nearly every patient

Evolutionary analysis revealed convergent and divergent paths to CDK4/6i + anti-ER treatment resistance in tumors with distinct lineages

Response to combined palbociclib, everolimus, and exemestane was correlated with activation of the mTOR pathway

Latest Intelligence

Unearthing New GEMS: Genesis and Gilead Collaborate to Tackle Tough Drug Targets

ESMO24: A cancer cachexia drug, VEGF enthusiasm and immunotherapy’s survival impact

The essential biotech updates of the day

Eli Lilly Sells Automated Lab to Arctoris, Platform Relocates to Oxford

Vironexis’ AAV vector approach: The next frontier in cancer immunology?

House members seek to block new wildlife refuge rules regarding GE crops

House hearing delves into antitrust problems of PBMs

Accelerating the new wave of CAR T therapy trials

How AI Makes Telehealth for STIs Smarter and More Accessible

The genome sequence of the Violet Carpenter Bee, Xylocopa violacea (Linnaeus, 1785): a hymenopteran species undergoing range expansion – Heredity