Polymorphism at codon 31 of CDKN1A (p21) as a predictive factor for bevacizumab therapy in glioblastoma multiforme

The study protocols were approved by the Medical Ethics Committee of Taichung Veterans General Hospital (Approval number: CF17263B-4). This retrospective study included 139 GBM patients ranging in age from 20 to 92 years. GBM diagnosis was confirmed through pathological examinations. Samples were obtained from the Department of Minimally Invasive Skull Base Neurosurgery, Neurological Institute, Taichung Veterans General Hospital, from 2010 to 2022, comprising the primary study cohort. Informed consent was obtained from all subjects or their legal guardians prior to surgery, and the collected samples were promptly frozen. The informed consent process involved providing detailed information about the study's objectives, procedures, potential risks, and benefits to the participants. They were given ample opportunity to ask questions and clarify any concerns before voluntarily providing their consent to participate. The study protocol and informed consent procedure were reviewed and approved by the relevant institutional ethics committee to ensure compliance with ethical guidelines and standards. All GBM patients underwent surgical resection and received concurrent chemoradiotherapy with temozolomide (Temozolomide (TMZ): 75 mg/m2/d) (CCRT), followed by adjuvant TMZ (150–200 mg/m2/d). Bevacizumab (10 mg/kg intravenously every 2 weeks until disease progression) was administered only to patients with recurrent GBM. The validation cohort consisted of 139 cases selected from the primary cohort based on the following criteria: (1) availability of follow-up data and samples, and (2) a post-operative survival time of more than one month. The obtained samples were immediately frozen after surgery. Overall survival (OS) time was defined as the time from the operation to the date of death or censored at the date of the last follow-up examination. The study end date was 31 March 2023. Commencing on May, 2012, the National Health Insurance Bureau of Taiwan broadened its coverage within the framework of health insurance benefits, encompassing the targeted drug bevacizumab's application in the treatment of adult patients experiencing relapsed glioblastoma multiforme. The administration of bevacizumab was typically synchronized with the emergence of disease progression, reflecting a commitment to transparency. This clinical determination was rooted in a thorough assessment of each patient's unique circumstances, alongside pertinent clinical variables. This systematic approach facilitated the precise deployment of bevacizumab, aligning its usage with the unmistakable signs of disease advancement. As a result, its focused deployment effectively managed the recurrence of glioblastoma. Nevertheless, it is imperative to acknowledge that certain individuals, prior to 2012, might not have received optimal counsel and treatment due to factors such as financial limitations or individual considerations.

The primary endpoint of this study focused on overall survival (OS), which was defined as the period in months starting from the initiation of the first surgery and extending until the time of death. In cases where patients were still alive at the point of data censoring, the OS was calculated up to the date of the last follow-up. The secondary endpoints encompassed progression-free survival (PFS) and the assessment of adverse events. PFS, specifically, was defined as the duration in months commencing from the initiation of bevacizumab treatment and extending to the occurrence of disease progression or death. If patients were alive and had not encountered disease progression during data censoring, the calculated time interval was extended to the date of the last follow-up. These endpoint definitions were selected to comprehensively evaluate treatment outcomes and patient experiences.

We adhered to SNP names in accordance with the guidelines set forth by the Human Gene Nomenclature Committee (HGNC). Genomic DNA was extracted from frozen tumor tissues for the genotyping assay. The CDKN1A c.93C > A (codon 31) polymorphisms were analyzed using a polymerase chain reaction (PCR)—restriction fragment length polymorphism (RFLP) assay. PCR–RFLP analysis is a rapid and straightforward technique employed as an additional method to detect genetic polymorphisms in GBM. However, this method has certain limitations. The sequences of partial CDKN1A c.93C > A patients were determined using a DNA autosequencer (GeneAmp PCR System 2700 Thermal cycler; Applied Biosystems) (Fig. 1b). The primer sequence and PCR conditions for CDKN1A c.93C > A are described in Table 1 and Fig. 1a. For each sample, the amplified PCR product was digested with the Blp I restriction enzyme (New England Biolabs, Beverly, Massachusetts, USA). The digested reactions were incubated for 16 h at 37℃ [23]. Subsequently, the genotyping assay was conducted on a 2% agarose gel using molecular weight markers and visualized after staining with ethidium bromide (Fig. 2a). The Ser allele harbors a single Blp I restriction site (GCTNAGC), resulting in two fragments of 89 bp and 183 bp, while the Arg allele remains undigested, yielding a single band of 272 bp (Fig. 2b). Each genotyping assay included positive and negative controls, and 10% of the samples were randomly selected and run in duplicates, showing 100% concordance. The results were reproducible with no discrepancies in genotyping. The Supplementary Materials contain the provided DNA sequencing data. Moreover, the heterozygote C/A genotyping of CDKN1A c.93C > A exhibited two signal peaks in the DNA sequencing data (Fig. 1b), consistent with expectations. Additionally, we analyzed two other genotypes of CDKN1A polymorphisms (CDKN1A c.168 + 16G > C, rs3176352G > C, IV2 + 16; CDKN1A 3’UTR c.*70C > T, rs1059234C > T, C70T) and found them to be highly linked with S31R (CDKN1A c.93C > A; C98A, rs1801270) (Fig. 1c). Unprocessed images of the DNA electric gel are provided in the Supplementary Materials. Additionally, the Supplementary Materials provide detailed experimental procedures for Methylation-specific PCR and the identification of the IDH1 gene.

Demographic data were presented as patient counts (percentages) for categorical variables and were compared using either the chi-squared test or Fisher's exact test. The overall survival (OS) outcomes were estimated using the Kaplan–Meier method, and differences in survival were assessed using the log-rank test. To investigate the association between CDKN1A and the 2-year overall survival (OS) of GBM patients, as well as the cumulative impact of CDKN1A SNP on the 2-year OS of GBM, adjusted hazard ratios (aHRs) with 95% confidence intervals (CIs) were employed. These aHRs were calculated using Cox proportional-hazards analyses and were adjusted for all the previously mentioned patient-level factors. Statistical analysis was conducted employing the Statistical Package for the Social Sciences (IBM SPSS version 22.0). All statistical analyses adhered to a two-sided approach, and significance levels of p < 0.05 and p < 0.01 were considered statistically significant.

This retrospective study included 139 glioblastoma patients, consisting of 58 males and 81 females. The average age of the participants was 56 years (range: 20–92). There were no statistically significant differences in gender distribution among the different genotypes of CDKN1A c.93C > A (codon 31) (p = 0.888). At the end of the study, 80.6% of the patients had passed away, while 19.4% were still alive, with a median survival of 16.8 months. The clinical characteristics of the patients and the associations between the utilization of bevacizumab and various patient characteristics are outlined in Table 2. It is worth noting that patients over the age of 70 generally exhibited a poorer prognosis after surgery. The different variants of CDKN1A c.93C > A polymorphisms did not show any significant correlations with patient age, gender, tumor number, tumor size, tumor occurrence, or response to bevacizumab treatment.

Association between CDKN1A c.93C > A Polymorphism and Glioblastoma Risk.

Tables 2 and 3 present the frequencies of genotypes and alleles within the CDKN1A gene. In our glioblastoma cases, we observed frequencies of 23.02% (32/139) for Ser/Ser, 27.34% (38/139) for Arg/Arg, and 49.64% (69/139) for Ser/Arg genotypes. Notably, all observed results were found to conform to the principles of the Hardy–Weinberg equilibrium. Furthermore, an assessment of the relationship between the CDKN1A c.93C > A genotype and demographic as well as clinicopathological characteristics of GBM patients was conducted, encompassing factors such as IDH1 gene status and MGMT promoter methylation status (as demonstrated in Table 3). Tables 2 and 3 provide additional insight by demonstrating that specific potential risk factors associated with GBM, including age, gender, tumor count, tumor resection size, IDH1 gene status, and MGMT gene methylation status, did not exhibit significant associations with the CDKN1A genotype. We investigated the link between individual CDKN1A c.93C > A genotypes and their correlation with the 2-year overall survival (OS) among GBM patients. This evaluation was conducted using Cox proportional-hazards models, which were adjusted for variables such as age, gender, stage, and the usage of bevacizumab, as illustrated in Table 4. However, in both univariate and multivariate analyses, none of the computed hazard ratios achieved statistical significance.

The study encompassed the enrollment of 139 GBM patients, with a median follow-up duration spanning 18.7 months (as visually represented in Fig. 3a). Our investigation ventured further into a meticulous assessment of the implications associated with the amalgamation of bevacizumab and chemotherapy concerning the outcomes of progression-free survival (PFS) and overall survival (OS) within the cohort of GBM patients. Within our study, we operationalized "PFS" as the timeframe commencing from the commencement of bevacizumab treatment and extending to the occurrence of disease progression or mortality. This particular definition was meticulously selected to exclusively probe the influence of bevacizumab treatment on the twin facets of disease progression and survival outcomes, with an explicit focus on the temporal period post-administration of the treatment. Although the median PFS duration within the subset of patients subjected to CCRT plus bevacizumab treatment did not manifest a statistically significant expansion (as visually indicated in Fig. 3b) (median of 14.5 months), our scrutiny of OS outcomes divulged a notably constructive impact. This was evident in the comparative evaluation between individuals who exclusively underwent chemotherapy (N = 69) and those who underwent a combined therapeutic regimen encompassing bevacizumab (N = 70), as vividly depicted in Fig. 3c.

Statistical analysis demonstrated a significant improvement (log-rank p < 0.001) in median OS from 9.7 to 28.2 months for the CCRT plus bevacizumab group compared to the CCRT group (Fig. 3c). These findings from the retrospective study strongly suggest that bevacizumab can extend the OS of patients with recurrent GBM [24].

To further explore the relationship between the CDKN1A c.93C > A polymorphism and GBM, we analyzed the overall survival data and genotyping information of all GBM patients. Using the log-rank test and Kaplan–Meier survival curve analysis, we found that the OS analysis comparing the CDKN1A c.93C > A variants did not yield significant results (Fig. 4a). However, a slightly better survival rate was observed in patients with the AA (Arg/Arg) variant.

We also investigated the impact of combining bevacizumab with chemotherapy on the overall survival (OS) of GBM patients with different CDKN1A genotypes. As shown in Fig. 4b, patients with the CC (Ser/Ser) genotype who received CCRT plus bevacizumab had a median survival of 17.3 months, similar to those with CCRT alone (p = 0.833). Interestingly, in contrast, patients with the AA (Arg/Arg) and CA (Ser/Arg) genotypes exhibited significantly longer median survival when treated with CCRT plus bevacizumab (34.3 and 24.9 months, respectively) compared to CCRT alone (13.1 and 8.5 months, respectively) (p = 0.001 and p < 0.001, respectively, Fig. 4c and d). Furthermore, patients with the CC genotype who received CCRT alone exhibited a higher median survival compared to the other two genotypes. Overall, GBM patients with the AA (Arg/Arg) and CA (Ser/Arg) genotypes demonstrated significantly prolonged survival in the CCRT plus bevacizumab treatment group compared to those with the CC (Ser/Ser) genotype (Fig. 4).

These findings suggest that GBM patients with the AA (Arg/Arg) and CA (Ser/Arg) genotypes of CDKN1A c.93C > A have significantly longer overall survival intervals when treated with CCRT plus bevacizumab compared to those with the CC (Ser/Ser) genotype in the same treatment group.

We broadened our inquiry to delve into the relationship between CDKN1A c.93C > A and the 2-year overall survival, stratifying the data according to the methylation status of the MGMT promoter and the IDH1 gene status (as illustrated in Table 5). Despite conducting both univariate and multivariate analyses, none of the calculated hazard ratios achieved statistical significance. Finally, we conducted a comprehensive risk assessment to determine the potential survival benefits associated with the utilization of BEV. Drawing from our research results, we employed the CDKN1A SNP, IDH1 gene status, and MGMT promoter methylation level to categorize risks. A summary of these findings is presented in Table 6. Among GBM patients with the CDKN1A c.93C > A genotype polymorphism, both univariate and multivariate analyses unveiled a substantial escalation in the risk of mortality for individuals with AA or CA genotypes who did not use BEV. In contrast, patients with the CC genotype exhibited no notable association with BEV usage. A comparable scenario is also evident within the MGMT methylation and IDH1 gene mutation groups: irrespective of MGMT methylation or IDH1 mutation status, individuals who refrain from BEV usage face a notably elevated risk of mortality compared to patients who undergo BEV treatment (Table 6). The aforementioned findings collectively suggest that the utilization of BEV appears to confer survival advantages to GBM patients, with the exception of those with the CDKN1A c.93C > A CC (Se/Ser) genotype.

This study exclusively involved GBM patients in Taiwan, making it particularly relevant to individuals with Asian heritage and nationality. To comprehensively explore the diversity of CDKN1A c.93C > A polymorphisms across different ethnicities, we conducted an extensive review of pertinent literature within the Asian population (references [19, 25‐28]) and the Caucasian population documented in the PubMed database (references [17, 29‐31]) (as illustrated in Fig. 5). While our study did not include a healthy control group, we meticulously analyzed relevant studies that provided accessible CDKN1A c.93C > A data in Asians, even if derived from different sources. Notably, these analytical samples were drawn from a variety of studies, albeit without distinguishing among different disease types.

As depicted in Fig. 5a, the distribution range for the Ser/Ser genotype of CDKN1A c.93C > A in the Asian population ranged from 21% to 45.83% (blue color), while the Arg/Arg and Arg/Ser genotypes had ranges of 17.97% to 29.21% (orange) and 33.33% to 54.5% (gray), respectively. Our data align with the expected distribution of CDKN1A c.93C > A genotypes in the Asian population (Fig. 5a, lane 1). Interestingly, among the Caucasian population (Fig. 5b, blue color), the distribution range for the Ser/Ser genotype of CDKN1A c.93C > A extends from 79.71% to 90.8%. Additionally, the distribution of the Ser/Arg genotype (gray color) ranged from 8.59% to 17.39%. The Arg/Arg genotype (orange color) was rarely found in the Caucasian population, with a distribution range of 0.1% to 1.26% (Fig. 5b). These findings highlight substantial differences in the distribution of CDKN1A c.93C > A genotypes between the Asian and Caucasian populations. However, it is important to note that these data alone do not establish CDKN1A c.93C > A as a risk factor for glioma.