This article is mentioned in:

Introduction

Patients with metastatic urothelial carcinoma (mUC) have a poor prognosis, with a 5-year survival rate of ~15% using conventional treatment regimens (1,2). The current standard treatment for patients with mUC, cisplatin-based chemotherapy, has a long history for mUC treatment but median overall survival (OS) times for patients with mUC remain at 14–15 months (3). Patients demonstrating mUC recurrence after first-line treatment, especially those with progression during first-line treatment, have a particularly poor prognosis. The introduction of immune checkpoint inhibitors (ICIs) as a second-line treatment has resulted in longer durations of treatment response and improved prognoses, but only for responders; the proportion of those patients is limited to 15–30% (4–6). Therefore, a non-invasive biomarker to evaluate response to ICIs is urgently required.

Regarding tissue-based biomarkers for ICI responses, several studies have reported the significance of tumor programmed death-1 (PD-L1) expression, tumor mutation burden (TMB) and microsatellite instability (MSI) (7–9). However, tissue biopsies present several problems in terms of invasiveness, accessibility and frequency of use in contrast to blood-based biomarkers obtained by liquid biopsy (10). Of note, plasma circulating tumor DNA (ctDNA) from liquid biopsies has shown particular promise in screening patients who may benefit from ICIs (11). In addition, ctDNA is useful for detecting minimal residual disease, monitoring disease status during treatment and predicting response to therapeutic drugs with a significant lead time to radiographic imaging (12–14). However, the significance of plasma ctDNA as a biomarker to evaluate pembrolizumab response in patients with mUC has not yet been fully investigated.

Therefore, the present study investigated the association between plasma ctDNA analysis and response to pembrolizumab in patients with mUC.

Materials and methods

Patients and sample collection

A total of 25 plasma samples from 16 patients with mUC were prospectively collected using PAX gene circulating free DNA (cfDNA) tubes (PreAnalytiX GmbH) at the University of Tsukuba Hospital (Tsukuba, Japan) between February 2018 and April 2021. Of these 25 samples, 16 were collected prior to administration of pembrolizumab (used as a second-line therapy) and nine were collected after administration of pembrolizumab in conjunction with computed tomography evaluation of clinical response. All patients provided informed, written consent to participate in the present study and the protocol was approved by the University of Tsukuba Hospital Institutional Review Board (Tsukuba, Japan; approval no, #H28-104). Clinical data were obtained from hospital charts. Response to pembrolizumab was evaluated according to the Response Evaluation Criteria in Solid Tumor (RECIST) guidelines (version 1.1) (15). In the present study, response to pembrolizumab was classified as complete response (CR), partial response (PR) and stable disease (SD).

Survival analysis

For analysis of OS, death from any cause was defined as an event and patients without any events were censored at the last follow-up visit. For progression-free survival (PFS) analysis, progressive disease (PD) evaluated by RECIST was defined as an event. The number of months from the date of administration of pembrolizumab to the event or censored date was calculated for OS and PFS.

Sample preparation and targeted sequencing

The extracted plasma cfDNA was subjected to a ctDNA analysis using the Avenio ctDNA Expanded Kit (Roche Diagnostics) designed to detect 77 genetic alterations, including single nucleotide variants (SNVs), indels and copy number variants (CNVs), according to the manufacturer's instructions. DNA concentration was assessed using a Qubit fluorometer (Thermo Fisher Scientific, Inc.). Prepared libraries were sequenced using a NextSeq 500 (Illumina, Inc) and analyzed using Roche AVENIO ctDNA Analysis Software (version 2.0.0; Roche Sequencing Solutions). Somatic variants were called and filtered with human reference genome hg38 using the same software and default logical and operation filter sets according to the manufacturer's instructions. Patients who carried ≥2 alterations (SNV, indels or CNVs) were defined as ctDNA-positive.

Statistical analysis

All statistical analyses were performed using R (version 4.0.2; R Foundation) and GraphPad Prism (version 8; GraphPad Software; Dotmatics). Kaplan-Meier survival curves were generated and the differences between the curves were compared with the log-rank test. For continuous variables, differences between groups were compared using the Mann-Whitney U-test. P<0.05 was considered to indicate a statistically significant difference.

Results

Patient characteristics

The characteristics of patients with mUC (n=16) treated with second-line pembrolizumab are presented in Table I. The median age of the patient pool was 70.0 years (range, 51–81 years) and the majority were male (75.0%). The most common histology was UC (n=13), followed by UC with small cell component (n=1) and UC with sarcomatoid variant (n=1). In terms of tumor location, 10 (62.5%) patients had bladder tumors and six (37.5%) had upper urinary tract tumors. Half of all patients presented with visceral metastasis.

Table I. Patient characteristics (n=16).

Table I.

Patient characteristics (n=16).

Characteristic	Value
Age, years	70.0 (51–81)
Sex
Male	12 (75.0)
Female	4 (25.0)
Follow-up time, months	9.5 (1.0–24.7)
Histology
UC	13 (81.3)
UC with small cell component	2 (12.5)
UC with sarcomatoid variant	1 (6.2)
Chemotherapy regimen
GC	9 (56.3)
GCa	4 (25.0)
GC+GCa	2 (12.5)
EP	1 (6.2)
Performance status
0	7 (43.8)
1	4 (25.0)
NA	5 (31.2)
Clinical response
PR	4 (25.0)
SD	5 (31.3)
PD	7 (43.7)
Tumor location
Bladder	10 (62.5)
Upper urinary tract	6 (37.5)
Visceral metastasis
Present	8 (50.0)
Absent	8 (50.0)

[i] Values are expressed as n (%) or the median (range). UC, urothelial carcinoma; GC, gemcitabine and cisplatin; GCa, gemcitabine and carboplatin; EP, etoposide and cisplatin; NA, not available; PR, partial response; PD, progressive disease; SD, stable disease.

As first-line chemotherapy before pembrolizumab administration, nine (56.3%) patients received a combination therapy of gemcitabine and cisplatin (GC), four (25.0%) patients received a combination therapy of gemcitabine and carboplatin (GCa), and one (6.2%) patient received a combination therapy of etoposide and cisplatin (EP). The remaining two patients (12.5%) received GC followed by GCa therapy as first- and second-line chemotherapies. Most patients (68.8%) demonstrated a good initial performance status of 0–1 at the time of pembrolizumab administration and second-line therapy resulted in nine (four PR and five SD; 56.3%) responders and seven non-responders (PD; 43.8%).

Genomic alteration profiles in plasma ctDNA and clinical outcomes

The genomic alteration profiles in plasma ctDNA collected from the 16 patients at baseline were analyzed. Unfiltered and filtered somatic variant data for each patient were collected. In total, 43 somatic variants (35 SNVs, one indel and seven CNVs) were identified with a median VAF of 2.0% (range, 0.08–55.8%) (Table SI). The observed genomic alteration profiles of analyzed plasma ctDNA were examined (Fig. 1). The median number of genomic alterations was 2.5 (range, 1–6) and all patients demonstrated ≥1 alterations. The most commonly altered gene was TP53 (75%), followed by ERBB2 (25%), PTEN (25%) and EGFR (19%). In terms of TP53 mutations, the most frequent subtype was missense (67%), followed by nonsense (25%) and splicing mutations (8%). Among ctDNA-positive patients, defined as those who demonstrated ≥2 genomic alterations, 10/11 (91%) had TP53 SNVs. There were no significant differences in cell-free DNA concentration after stratification by ctDNA positivity (Fig. S1). The characteristics of patients with mUC were also stratified by ctDNA status (Table II). The neutrophil-lymphocyte ratio of patients in the ctDNA-negative group was significantly higher than that in the ctDNA-positive group (7.28 vs. 2.65).

Figure 1. Genomic alteration profiles of plasma ctDNA analyzed by targeted sequencing. Patients who demonstrated ≥2 alterations were defined as ctDNA-positive. ctDNA, circulating tumor DNA; SNV, single nucleotide variant; CNV, copy number variant; PD, progressive disease; PR, partial response; SD, stable disease.

Table II. Patient characteristics stratified by ctDNA status.

Table II.

Patient characteristics stratified by ctDNA status.

Characteristic	cthNA-positive (n=11)	cthNA-negative (n=5)
Age, years	71.0 (59–81)	69.0 (51–74)
Sex
Male	8 (72.7)	4 (80.0)
Female	3 (27.3)	1 (20.0)
Follow-up time, months	10.4 (1.0–24.7)	5.6 (2.4–11.9)
Chemotherapy regimen
GC	6 (54.5)	3 (60.0)
GCa	2 (18.2)	0 (0.0)
GC+GCa	2 (18.2)	2 (40.0)
EP	1 (9.1)	0 (0.0)
Performance status
0	6 (54.5)	1 (20.0)
1	2 (18.2)	2 (40.0)
NA	3 (27.3)	2 (40.0)
Tumor location
Bladder	8 (72.7)	2 (40.0)
Upper urinary tract	3 (27.3)	3 (60.0)
Visceral metastasis
Present	5 (45.5)	3 (60.0)
Absent	6 (54.5)	2 (40.0)
WBC count	6,000 (2,900–9,700)	8,900 (3,600–22,400)
Neutrophil count	3,830 (1,302–6,383)	6,602 (2,045–20,026)
Lymphocyte count	1,428 (378–2,408)	907 (417–1,958)
NLR	2.7 (1.3–12.1)	7.3 (1.7–40.0)a

a P<0.05 vs. ctDNA-positive group. Values are expressed as n (%) or the median (range). ctDNA, circulating tumor DNA; GC, gemcitabine and cisplatin; GCa, gemcitabine and carboplatin; EP, etoposide and cisplatin; NA, not available; WBC, white blood cell; NLR, neutrophil-lymphocyte ratio.

Association between mutation frequency data from target sequencing and response to pembrolizumab

The association between mutation profiles and pembrolizumab response was analyzed. The proportion of responsive patients in the ctDNA-positive group was higher than that in the ctDNA-negative group (72.7 vs. 20.0%) (Fig. 2A). After stratification by ctDNA positivity, median response times in the ctDNA-positive group were significantly longer compared with those in the ctDNA-negative group (3.2 vs. 0 months) (Fig. 2B). Furthermore, 4/11 (36.4%) patients in the ctDNA-positive group demonstrated a sustained response at data cutoff, while none in the ctDNA-negative group had such a response.

Figure 2. Proportions of (A) patient responses and (B) response statuses in individual patients over time stratified by ctDNA status. ctDNA, circulating tumor DNA; PR, partial response; SD, stable disease; AE, adverse event.

The association between ctDNA positivity and patient prognoses, as well as response to pembrolizumab, was analyzed. The median PFS of patients in the ctDNA-positive group was significantly longer compared with that in the ctDNA-negative group (3.9 vs. 2.0 months) (Fig. 3A) and the median OS demonstrated a similar trend (14.7 vs. 5.6 months) (Fig. 3B). These results suggested that the mutation frequency in plasma ctDNA was associated with clinical response to pembrolizumab and disease outcomes.

Figure 3. (A) Progression-free survival and (B) overall survival of patients treated with pembrolizumab stratified by ctDNA status. + in the Kaplan-Meier curves indicate censored data-points. ctDNA, circulating tumor DNA.

Association between changes in the VAF of each mutation in plasma ctDNA and pembrolizumab response

The association between mutation frequency and clinical response during pembrolizumab treatment using samples from eight (50.0%) of 16 patients collected during pembrolizumab treatment was examined. There was no association between changes in the mutation frequency and individual clinical responses (Fig. S2). Next, the association between changes in the VAF of each mutation and individual clinical responses was examined. The UC03 patient who demonstrated PR with shrinking lymph node metastasis had decreased VAF in TP53, RB1 and KDR, whereas the VAF in ERBB2 and RAF1 was increased (Fig. 4A). Another patient (UC22), who demonstrated PD with enlargement of lymph node metastasis, had increased VAF in TP53, AKL and GNA11 and decreased VAF in PTEN (Fig. 4B). Among these mutated genes, only the VAF change in TP53 was associated with clinical response. In terms of the VAF changes for each mutated gene in the other six patients, only TP53 was associated with clinical response during pembrolizumab treatment (Fig. 4C). These results suggested that the VAF changes in TP53 during pembrolizumab treatment accurately reflected the therapeutic response.

Figure 4. Changes in variant allele frequency for each mutation in eight patients with (A) PR, (B) PD and (C) other disease statuses. Computed tomography images indicate changes of lymph node metastasis (red arrows). Samples were collected up to three times during pembrolizumab treatment. PD, progressive disease; PR, partial response; VAF, variant allele frequency; ND, not detected.

Discussion

The present study, using genomic alteration profiling of plasma ctDNA collected from 16 patients with mUC at baseline, demonstrated that ctDNA-positive status was associated with patient response to pembrolizumab treatment. Furthermore, analysis of consecutive samples from eight patients during treatment demonstrated that decreases in TP53 VAF accurately reflected therapeutic response. These results suggested that profiling of plasma ctDNA was useful in predicting and monitoring the therapeutic pembrolizumab response in patients with mUC.

Blood-based biomarkers, which may be detected by liquid biopsy in a non-invasive manner, are useful in screening responsive patients (11); however, there are no currently established and approved blood-based biomarkers for predicting responses to ICI. While TMB is recognized as a tissue-based biomarker for ICI response (8,16), previous studies suggest that elevated TMB in blood, a liquid biopsy-style marker as detected by ctDNA assays, is also associated with responses and prognoses in patients treated with ICI (17–19). Si et al (18) previously reported that blood TMB 320 mutations/megabase was useful in predicting clinical benefit from ICI treatment compared with chemotherapy. In addition, Kim et al (19) reported that ctDNA mutational load scores correlate with clinical pembrolizumab responses in metastatic gastric cancer. Although the present study did not examine the blood TMB, an association between ctDNA-positive status and ICI response was demonstrated. On the other hand, Laukhtina et al (20) reported an association between high baseline ctDNA levels and worse disease-free survival and OS. This may be due to different profiling of the mutated genes, including the percentage of TP53 mutations, between the previous studies and the present study. A number of studies previously reported that TP53 mutations are associated with superior clinical benefits to ICI treatment (21–23). In line with these studies, the present study demonstrated that TP53 mutations were associated with an improved patient response to ICI. Sun et al (22) reported that TP53 missense mutations were associated with better clinical benefit in their local clinical cohort, but nonsense mutations were not. Similar trends were also reported in the MSK study cohort, while the Checkmate-012 study cohort reported an opposite trend (22). In the present study, TP53 missense mutations were the most common subtype of TP53 mutations.

A recent study reported that ctDNA clearance during ICI treatment in advanced urothelial carcinoma is associated with better clinical outcomes (24), which led the present study to investigate VAF changes for each mutation during second-line pembrolizumab treatment. The tumor suppressor TP53 is the most frequently mutated cancer-associated gene in urothelial bladder cancer (25), a result also demonstrated by the present study. Thorsson et al (26) suggested that driver mutations, such as TP53, by inducing genomic instability, may alter the immune environment through the production of neoantigens and paradoxically produce better response to ICIs. However, among the genomic alterations detected in the present study, decreases in TP53 VAF were associated with improved clinical response during pembrolizumab treatment. Similarly, Parkinson et al (27) reported that, in high-grade serous ovarian carcinoma, TP53 VAF changes during chemotherapy correlated with clinical response. The findings of the present study, which suggested that decreases in TP53 VAF during pembrolizumab treatment were associated with therapeutic response, open up new possibilities for frequent liquid biopsies to screen first-line non-responders for second-line feasibility.

The present study has several limitations. First, the number of patients studied was small, so larger patient cohorts are necessary for future studies. Second, technological restrictions limited the assessable genes to 77; therefore, the blood TMB and MSI status of samples could not be identified. Third, as the study protocol did not include any tissue analysis, the TMB and PD-L1 expression could not be evaluated, both of which have been previously reported to correlate with ICI response (7–9). Studies have reported that sequential ctDNA analysis can identify responders to ICI therapy faster when compared with radiographic imaging (12,13). As the present study protocol did not determine the timings of liquid biopsies and radiographic evaluation, it was difficult to assess the time saved for ctDNA analysis over radiographic imaging. Finally, variants associated with clonal hematopoiesis of indeterminate potential could be mistaken for tumor-specific variants.

In conclusion, plasma ctDNA analysis in patients with mUC treated with pembrolizumab demonstrated an association of TP53 mutation frequency with clinical response of patients. Furthermore, decreases in TP53 VAF during pembrolizumab treatment was associated with clinical response. These results suggest that plasma ctDNA analysis of patients with mUC could potentially be used for predicting both pembrolizumab response and monitoring disease status.

Supplementary Material

Supporting Data

Supporting Data

Acknowledgements

The authors would like to thank Mr. Noriyo Ishibashi (Department of Inspection, Tsukuba i-Laboratory LLP), Mrs. Noriko Kunita (Technical Assistant, Department of Urology, University of Tsukuba) and Mrs. Naoko Ueki (Technical Assistant, Department of Urology, University of Tsukuba) for their technical support in sample processing and library preparation.

Funding

This work was supported by a grant from the Japan Society for the Promotion of Science KAKENHI (grant no. 21K09365).

Availability of data and materials

Sequence data generated during the current study are available in the Japanese Genotype-phenotype Archive (https://ddbj.nig.ac.jp/resource/jga-study/JGAS000622). The generated somatic variant data for each patient is presented in Table SI.

Authors' contributions

YN, TK and BJM contributed to the study design and technical and material support. KH, YN, SK, KT, MS, AH, HNe and HNi performed sample preparation. KH, YN and TK analyzed the acquired data. KH and YN drafted the manuscript. KH, YN, SK, BJM, HNe and HNi revised the manuscript critically. KH and YN confirm the authenticity of all the raw data. All authors have read and approved the final manuscript.

Ethics approval and consent to participate

The study protocol and data processing were approved by the University of Tsukuba Hospital Institutional Review Board (approval no. H28-104). Written informed consent was obtained from all patients involved in the study.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References