Prostate cancer (PC) remains the most diagnosed malignancy and the second leading cause of cancer-related death among men in the United States. Between 2014 and 2019, prostate cancer incidence has increased by 3% per year (1, 2). Although most men diagnosed with PC have local disease, recurrences frequently occur (3, 4). Furthermore, diagnosis of metastatic prostate cancer (mPC) has been modestly increasing in the context of changing screening practices and is a dynamic disease with a poor prognosis, particularly as it progresses to castration-resistant prostate cancer (CRPC) (5–8). Prostate cancer has the propensity to develop resistance across various treatment paradigms, which include androgen axis targeting agents and chemotherapy. Thus, uncovering the genomic phenotypes as the disease evolves is critical to understanding patient prognosis, treatment sequencing, and developing novel therapeutics. Circulating tumor DNA (ctDNA) provides an essential tool for understanding such genomic phenotypes, guiding therapeutic decisions, and characterizing tumor heterogeneity in a cost-effective manner. It poses minimal patient risk and little technical challenge compared to (repeated) biopsy of tumor tissue (8–10).

Recently, the use of ctDNA analysis has increased, enhancing our understanding of PC progression and helping to guide clinical management (9). An analysis of more than 500 patients with metastatic CRPC (mCRPC) identified frequent androgen receptor (AR) and MYC alterations associated with clinical outcomes, such as overall survival (OS) and failure-free survival (FFS) (11). Several AR mutations that are associated with resistance to anti-androgen axis therapies were identified by ctDNA analysis and can guide therapeutic decision-making (12). Although distinct from ctDNA, circulating cell-free DNA (cfDNA), which consists of ctDNA and hematopoietic cell-derived DNA, is elevated in patients with mPC and can also serve as a marker of therapeutic response. For example, decreased cfDNA levels were associated with improved outcomes in patients harboring mutations in genes involved in homologous recombination and demonstrating clinical response to poly(ADP)ribose polymerase (PARP) inhibitor (13). Similarly, cfDNA levels were reduced in patients receiving taxane therapy, and that reduction was associated with both radiological progression-free survival (rPFS) and OS (14). Finally, ctDNA levels were associated with rPFS in mCRPC patients in the A.MARTIN phase II study of abiraterone with or without pan-AKT inhibitor ipatasertib (15).

In patients with mPC, the genomic landscape remains fluid throughout the treatment course. ctDNA can be used to profile somatic mutations and genomic amplifications (GAs) with their corresponding variant allele frequency (VAF) plasma copy number (PCN) during any stage of the treatment course. GAs may be representative of either chromosomal duplication and/or gene-specific amplification, which is particularly interesting because aneuploidy is a relevant feature of aggressive prostate cancers that may be more likely to be lethal (16). Copy number alterations frequently occur in the form of loss of regions, with the loss of tumor suppressors CHD1, RB1, TP53 and PTEN as the most common alterations (17). A recently published copy number alteration analysis from 300 patients in the androgen deprivation therapy control arm of the STAMPEDE trial showed that loss of segments of chromosome 5 containing CHD1 and amplifications in segments of chromosome 8 containing MYC were associated with a higher burden of copy number alterations. Copy number alteration burden was associated with a statistically significant increase in metastasis at diagnosis, risk of progression, and death (18). This follows prior demonstration of the presence of AR, or MYC, or BRAF amplifications (as detected by ctDNA) are associated with worse OS (11). Furthermore, we previously demonstrated that elevated carcinoembryonic antigen (CEA) is associated with higher rates of liver metastases and increased copy number alterations of select genes (19).

The prognostic significance of GAs and PCN detected prior to or serially through the course of treatment in a cohort of patients remains to be fully characterized. The majority of ctDNA assays do not commonly report on actionable deletions, germline mutations, or less frequent GAs such as chromosomal rearrangements (20). This limitation led us to focus on whether GAs and PCN measured through ctDNA analysis alone can inform us on clinical outcomes and therefore continue to build from prior studies. To investigate the associations between GAs and PCN with specific clinical factors and treatment outcomes, we conducted a retrospective analysis of patients at our institution with mPC who harbor GAs.

Circulating tumor DNA analysis is increasingly being used in mPC and its role in decision making is being defined. While many commercially available assays examine the presence of single nucleotide variants, few ctDNA assays report GAs. Our data suggest that clinically useful information may be found in the additional analytes. Furthermore, our findings provide a basis for evaluating ctDNA somatic alterations at diagnosis of metastatic disease and throughout the course of treatment, with an emphasis on identifying GAs and continually assessing changes in their PCN. This paradigm of serial ctDNA analyses might be used not only to guide therapeutic decisions but also inform prognosis and serve as a biomarker of treatment response, complementing known disease markers such as PSA. Of note, prior studies have evaluated early changes of ctDNA across solid tumors and found that patients who achieve ≥50% decrease in their ctDNA ratio early in immunotherapy-based treatment as compared to baseline sampling have a superior PFS and OS advantage compared to patients who do not reach a 50% threshold for decrease (24, 25). While our cohort does contain serial sampling, the timepoints of collection are diverse and are not reflective of an early on-treatment change. Thus, we aimed to demonstrate that any decrease in ctDNA or amplifications in patients with mCRPC, even at a low threshold of 10% reduction, is prognostic of clinical benefit to their therapy.

Previous studies have evaluated the clinical significance of specific mutations and the associations they may have in clonal development and treatment resistance (12–14). Although most patients in prior studies have had detectable co-existing mutations, our study emphasizes evaluation of clinical factors and outcomes in patients stratified based on specific GAs and changes in PCN in prostate cancer. Our approach is supported by the fact that GAs are commonly observed in patients with mPC, particularly when disease is castration-resistant. By identifying specific GAs, ctDNA analysis could prove a valuable management tool for patients seen at various points in their sequential treatment courses.

Inferior outcomes have been associated with alterations in AR, such as coding sequence mutations, splice variants, and amplifications (26). With regard to this study, it is important to note findings regarding presence of the number of GAs and which genes may be amplified are nuanced, and need to be evaluated in context (i.e. whether they co-occur with other GAs). For instance, prior work has demonstrated AR amplification being associated with reduced FFS and OS (11). However, in our study, “AR Plus” patients have worse OS compared to patients exclusively with an AR GA ( Figure 5 ). This demonstrates AR amplification in the presence of other GAs is associated with worse clinical outcomes. In other words, tumor biology in mCRPC may not be as aggressive in a patient with only an AR amplification compared to those whose malignancy has multiple GAs that include AR (such as MYC, BRAF, PIK3CA, and others).

Increased numbers of GAs are seen in patients with TP53 mutations ( Figure 2C ). This is consistent with TP53 mutations being associated with increased aneuploidy frequency in malignancies (27, 28). Previous studies have identified putative associations and roles for BRAF, PIK3CA, and FGFR1 in mCRPC (17, 29, 30). The association we found between BRAF or PIK3CA GAs, and liver or lung metastases, further prioritizes the need to identify potential treatment options for patients with these GAs ( Figure 2D ). This also corresponds to reduced OS seen in patients with PIK3CA amplification relative to other patients with GAs ( Table 2 and Figure 6 ). A recent report from our group demonstrated CEA elevation is associated with a number of amplifications including PIK3CA (19). However, this was not associated with aggressive variant prostate cancer phenotype (19). Further evaluation regarding the biological input of PIK3CA amplification is needed.

Our study identified a high frequency of BRAF amplifications compared to the TCGA database. We feel this reflects the advanced disease of the subjects evaluated in this study. As multiple genes are co-amplified with BRAF such as CDK6, EGFR, and MET (also located on chromosome 7), this gives confidence that we are in fact detecting true BRAF amplifications. Related to this, we found that CDK6 amplification was significantly associated with liver/lung metastases; however, it remains to be seen whether this association reveals a significant biological role for CDK6 in disease pathogenesis or treatment. Finally, we observed an association between FGFR1 amplification and liver/lung metastases ( Figure 2D ). Pre-clinical studies have demonstrated that FGFR1 is involved in prostate cancer progression and has a role in the metabolic reprogramming of prostate cancer cells (29, 30).

Serial ctDNA analyses revealed the emergence of GAs in multiple patients following systemic therapy ( Figure 8 , PtID’s 71, 101, 114, 135, and 175). In fact, some patients developed new GAs even after a reduction in AR and/or MYC PCN (i.e. PtID’s 18, 69, 87, and 125). The latter phenomenon is likely reflective of tumor heterogeneity, with certain clones/subclones sensitive to treatment resulting in PCN reduction, while new subclones then emerge that are resistant to treatment. In the copy number analysis performed on the STAMPEDE patient cohort, sequencing was performed on multiple regions of the same prostate tumor, and intratumoral differences in GA burden, indicating increased heterogeneity, was associated with increased metastatic potential (18).

This study has a number of limitations. It is a single-institution retrospective study intended to be hypothesis-generating in nature and as such does not capture the breadth of geographic differences in patient clinical accessibility, outcomes, and provider bias for ctDNA testing and treatment recommendations. Furthermore, this study had a modest sample size and does not represent the full extent of tumor genomic heterogeneity among patients. Although 155 patients with mPC were evaluated in this cohort, the survival analysis was limited to the 91 subjects in Group A and Group B, ( Figure 1 ), who had mCRPC and adequate treatment records for the entire disease course ( Figures 3 , 4 , 7 ). Provider bias in the use or sequencing of abiraterone could have affected our finding that it was prognostic for FFS. Although the gene panel used in this study is meant to capture a wide variety of somatic alterations relevant to numerous malignancies, it evaluates a modest number of genes, ranging from 68-74. With regard to mPC, this panel did not evaluate somatic alterations such as AR variant 7 splice site changes, SPOP mutations, and PTEN loss (31, 32). Therefore, these are beyond the scope of this study. Moreover, in the analysis completed in group B, there is variability among patients in the timing between ctDNA analyses. This variability also extends to the timing and variety of systemic treatments administered prior to initial AR/MYC GA detection and between ctDNA analyses. Future work in controlled trial formats should investigate PCN changes in relevant GAs.

Nonetheless, this study provides significant support for the wider use of ctDNA evaluation to guide prognosis and mPC treatment selection. Additionally, it reinforces the role for serial monitoring of ctDNA to characterize changes in GAs and other somatic alterations. Ultimately, this raises the question whether ctDNA information should be incorporated into decision making regarding early treatment options. For instance, should mPC patients receive more aggressive initial systemic therapy such as the addition of docetaxel to ADT plus androgen axis targeting therapies (i.e. darolutamide or abiraterone as per the ARASENS or PEACE-1 studies) if they harbor GAs that are linked to increased likelihood of visceral metastases (33, 34)? Ongoing prospective clinical studies are evaluating the impact of pretreatment ctDNA evaluation on therapeutic decisions and clinical outcomes in metastatic disease. Additionally, it remains unknown whether specific amplifications are associated with increased expression of their respective gene products and enhanced downstream signaling of relevant oncogenic pathways. If such an association exists, protein products of select GAs would make attractive therapeutic targets.

Findings from our retrospective study demonstrate that GAs, as detected by ctDNA analysis, may play a significant role in informing potential risk for visceral metastases and overall prognosis in mCRPC. Furthermore, evaluation of serial PCN during mCRPC treatment provides an additional tool to determine putative responses to a variety of therapeutic interventions and, as such, can supplement clinical decision-making. Future studies are needed to determine the prognostic significance of specific GAs and the reliability of using PCN as a clinical tool to assess treatment responses.

The datasets presented in this article are not readily available because of patient anonymity. Requests to access the datasets should be directed to the corresponding author.

The studies involving humans were approved by Medical University of South Carolina IRB. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and institutional requirements.

TD contributed to conceptualization, data analysis, study design, and was the primary manuscript writer/editor. ML contributed to conceptualization, data analysis, study design, and manuscript writing/editing. HL performed formal statistical analysis and manuscript review/editing. JK, LD, and AG contributed to data analysis, and manuscript writing/editing. TG and PH contributed to manuscript review/editing. All authors contributed to the article and approved the submitted version.