CYP3A5 Polymorphism in Circulating Tumor Cells Confers an Increased Disease-Free Survival in DLBCL Patients Treated with R-CHOP

Introduction

Diffuse Large B-Cell Lymphoma (DLBCL) is the most common aggressive type of Non-Hodgkin Lymphoma (NHL) in Mexico and worldwide.^1–3 The presence of Circulating Tumor Cells (CTCs) in peripheral blood, which derive from primary tumors, has been demonstrated to cause dissemination and spread of the disease to other organs and tissues. Our research group has identified and molecularly characterized CTCs in liquid biopsies from DLBCL patients.⁴ This methodology allows a faster and less invasive study of molecular alterations.

The current standard treatment for DLBCL continues to be the R-CHOP regimen (Rituximab-Cyclophosphamide, Hydroxydaunorubicin (Doxorubicin), Oncovin (Vincristine), and Prednisone), achieving response rates ranging from 50% to 75%.^5–9 Several molecular alterations can cause treatment failure, such as the presence of certain SNPs (Single Nucleotide Polymorphisms) in genes involved in drug metabolism, for example those of the cytochrome P450 and drug resistance genes (ABC/MDR-1).^10,11

Cytochrome P450 encoding genes include 18 families, divided into 41 subfamilies, encompassing 57 genes.¹² Members of the cytochrome P450 3A (CYP3A) family include CYP3A4, CYP3A5, CYP3A7, and CYP3A43 genes, which are involved in elimination and metabolism of drugs, xenobiotic compounds, and endogenous molecules.^13,14 For the CYP3A5 subfamily, 34 SNPs have been identified, one of the most studied being the A to G transition (A6986G) within intron 3. This SNP alters the correct splicing of mRNA transcription, leading to low levels of protein expression and, consequently, decreased enzymatic activity.^15,16

On the other hand, ATP-binding cassette (ABC) drug resistance genes are part of seven subfamilies ranging from A to G, with 49 genes identified.¹⁷ The B Subfamily consists of 11 genes, which confer resistance to multiple drugs in cancer.¹⁸ One of the most studied is the ABCB1 gene, which consists of 28 introns and 28 exons that encode a membrane transporter protein called P-glycoprotein (P-gp).¹⁹ More than 50 SNPs of the ABCB1 gene have been identified, being C3435T the most studied and addressed by our group. This variant is associated with decreased mRNA expression and protein stability. It has been described that this SNP is related to multidrug resistance in cancer and various pathologies such as epilepsy.^20–22 ABCG2 is another multidrug transporter gene that affects pharmacokinetics and contributes to the resistance of various types of cancers. Originally known as the Breast Cancer Resistance Protein (BCRP), it is located on chromosome 4q22 and encodes a 72 kDa membrane protein.^23,24 One of the most studied SNP in this gene is located in exon 5 (C421A), causing a substitution of glutamine by lysine, which leads to a decrease in the expression and activity of the protein. It has been demonstrated that CYP3A enzymes and the ABCB1 drug transporter share the same substrates.^14,25

Despite the development of therapeutic strategies with new molecules, cases of progression and relapse persist. Therefore, the aim of this study was to assess the clinical implications of these SNPs involved in drug metabolism in DLBCL patients. Identifying them at the beginning of treatment could allow their use as potential biomarkers to introduce these SNPs as new prognosis factors that predict the individualized behavior of the chemotherapy response.

Materials and Methods

Study Population

We included 102 liquid biopsies from patients with DLBCL at diagnosis, from the Hospital General de México “Dr. Eduardo Liceaga”, from January 2018 to December 2022. The included patients had a complete medical record and signed informed consent. The mean age at diagnosis of the study cohort was 53 years (range 18 to 82), with a predominance of females with 53 patients (52%). Among other relevant clinical parameters, the ECOG (Eastern Cooperative Oncology Group) performance status of 69 patients (67.6%) was between 0 and 1, 40 patients (39.2%) had Lactate Dehydrogenase (LDH) levels above the reference limit (271 U/L), 65 patients (63.7%) were diagnosed in advanced stages of the disease (III and IV), of which 39 patients (38.2%) had extranodal sites. Based on these parameters, 64 patients (62.7%) had an IPI score (International Prognostic Index) between 0 and 2 (Low, Low Intermediate) and 38 (37.3%) had a score between 3 and 5 (Intermediate High - High). Furthermore, according to the Hans algorithm (CD10, BCL6, and MUM1), the predominant cell of origin was non-Germinal Center B-cell (non-GCB) with 53 patients (52%), while 49 (48%) corresponded to Germinal Center B-cell (GCB) (Table 1).

Table 1 Clinicopathological Characteristics of DLBCL Patients (n=102)

Treatment Response

The first-line treatment regimen used was R-CHOP (Rituximab 375 mg/kg, Cyclophosphamide 750 mg/kg, Doxorubicin 50 mg/kg, Vincristine 1.4 mg/kg, and Prednisone 1 mg/kg) administered in 6 cycles, every 21 days.

During a median follow-up time of 855 days, a total of 60 (58.8%) patients completed their first-line treatment regimen with a complete response (CR), maintaining it throughout this study. On the other hand, 42 (41.2%) patients had an unfavorable clinical outcome, with 17 deaths (16.7%), of which 1 was due to relapse and 6 due to treatment refractoriness, and the rest (n=10) died before completing the treatment. Additionally, there were 7 (6.9%) relapse cases and 18 (17.6%) refractory cases, resulting in overall survival (OS, the percentage of patients that remained alive from DLBCL diagnosis to death or surveillance length) rate of 83.3% (85/102) and disease-free survival (DFS, the percentage of patients that remained free of disease complications, death, relapse progression through the time of study duration) rate of 58.8% (60/102).

CTCs Enrichment

From each patient, an 8 mL liquid biopsy of peripheral blood was obtained before the administration of the first R-CHOP cycle. The total blood volume was mixed with PBS 1X and Lymphoprep (Axis-Shield, Oslo, Norway), according to the manufacturer’s recommendations for isolating mononuclear cells and CTCs by density gradient. Subsequently, B cells from the liquid biopsy were isolated by positive selection using anti-CD20 microbeads (marker present in DLBCL cells), (Miltenyi Biotec, Bergisch Gladbach, Germany), according to the manufacturer’s recommendations.

DNA Extraction and Purification

DNA was extracted using DNAzol reagent according to the manufacturer’s recommendations (Invitrogen, Life Technologies Carlsbad, CA). Subsequently, DNA purification was performed using the QIAamp DNA Mini Kit (Qiagen, Maryland, US). The genetic material was stored at −20°C until used. The concentration and purity of the DNA were evaluated with a Nanodrop 2000 (ThermoScientific, Wilmington, DE, US).

Detection of C3435T, C421A and A6986G SNPs on CTCs

To detect the polymorphisms rs1045642 (C3435T), rs2231142 (C421A), and rs776746 (A6986G) of the genes ABCB1, ABCG2, and CYP3A5, respectively, we used SNP Genotyping Assay ABCB1 (C_7586657_20), ABCG2 (C_15854163_70), and CYP3A5 (C_26201809_30) (ThermoScientific, Wilmington, DE, US), TaqMan Genotyping Master Mix, and 10 ng of genomic DNA, in the Step One™ Applied Biosystems instrument.

Statistical Analysis

Pearson correlation analyses were performed to correlate clinical-pathological parameters (co-variables), treatment response/clinical outcome (dependent variables), and the genotypes of the analyzed genes (independent variables). For OS and DFS analysis, the Kaplan-Meier and Log Rank methods were used. Odds Ratio (OR) risk analysis was performed by Cox regression. A p-value less than 0.05 was considered statistically significant. All statistical analysis was performed using SPSS version 25 software (IBM, Armonk, NY, US).

The sample size (N=102) was calculated a priori using G*power 3.1.9.7 software, based on an effect size of 0.8, α value of 0.05, and 95% confidence level.

Ethical Considerations

This trial is a minimum risk investigation and was performed based on the Declaration of Helsinki. For its realization, it was approved by the Ethics and Research Committees of the Hospital General de México “Dr. Eduardo Liceaga” with registration numbers DI/19/103/03/00 and DI/16/103/03/035.

Results

Characteristics of the Cohort Study

The frequency of polymorphisms in DLBCL patients is shown in Table 2.

Table 2 ABCB1, ABCG2 and CYP3A5 Genotype Frequencies in DLBCL Patients

For the evaluation of clinically relevant prognostic variables, patients were divided into two groups based on clinical response: the first group of patients with a maintained complete response and the second group of patients with progression, relapse, or death.

For the first group (maintained response), there were 60 patients, with 24.5% (25/102) over 60 years old, 30.4% (31/102) were female and 28.4% (29/102) were male, 15.7% (16/102) had an ECOG score higher than 2, and 18.6% (19/102) had involved extranodal sites, 31.4% (32/102) were in stages III and IV vs 27.5% (28/102) in stages I and II, with an IPI score of 3–5 in 19.6% (20/102). From this group, 33.3% (34/102) of patients were classified as non-GCB.

The second group consisted of 42 patients with progression, relapse, or death. Of these, 10.8% (11/102) were over 60 years old, 21.6% (22/102) were female and 19.6% (20/102) were male, 18.6% (19/102) had elevated LDH values, 16.7% (17/102) had an ECOG score of 2 or higher, 19.6% (20/102) had extranodal sites, 17.6% (18/102) had an IPI score of 3–5. From this group, 18.6% (19/102) of patients corresponded to non-GCB cases. There were no significant differences between the two groups except for the clinical stage (p=0.009).

The frequency of the C3435T, C421A, and A6988G genotypes of the ABCB1, ABCG2, and CPY3A5 genes showed that in the group of patients with progression, relapse, and death, the CC/CT genotype was 31.4% (32/102) and the TT genotype was 9.8% (10/102) for the ABCB1 C3435T; the CA/CC genotype was 35.3% (36/102) and AA was 5.9% (6/102) for the ABCG2 C421A; and the AG/AA genotype was 25.5% (26/102) and GG was 15.7% (16/102) for the CYP3A5 A6986G.

In the group of patients with complete response, the CC/CT genotypes of the ABCB1 C3435T were 48% (49/102) and the TT genotype was 10.8% (11/102); the CA/CC genotypes of the ABCG2 C421A of the gene were 52.9% (54/102) and the AA genotype was 5.9% (6/102); and finally, for the CYP3A5 A6986G, the AG/AA genotypes were 24.5% (25/102) and the GG genotype was 34.3% (35/102), obtaining a statistically significant result (p=0.044) (Table 3).

Table 3 Correlation Between Clinical Parameters-Polymorphisms and Clinical Outcome in DLBCL Patients (n=102)

Polymorphisms and Overall/Disease-Free Survival

In an average follow-up time of 855 days, the OS was 83.33% and the DFS was 58.82% (Figure 1). Comparing the polymorphisms with OS, no significant results were found with any of them.

Figure 1 Overall Survival (OS) and Disease-Free Survival (DFS) of DLBCL patients (N=102). (A) General OS (83.3%). (B) General DFS (58.8%).

For DFS, the GG genotype of CYP3A5 A6986G showed longer survival of 68.6% with 1534 days (CI 1286–1782) vs the AG/AA genotypes, with 49% (1038 days, CI 812–1263) Log Rank p=0.019 (Figure 2A). No significant results were found with the polymorphisms of the ABCB1 and ABCG2 genes (Figure 2B and C).

Figure 2 DFS of DLBCL patients (N=102) according to genotype of polymorphisms. (A) CYP3A5 A6986G. (B) ABCB1 C3435T. (C) ABCG2 C421A.

Subsequently, DFS in patients with the GG genotype of CYP3A5 A6986G was evaluated by stratifying according to clinical parameters, finding a relationship with good prognostic variables: age equal to or less than 60 years (p=0.001, GG 68.6% vs AG/AA 35.5%); and normal LDH levels (N=62) (p=0.027, GG 74.2% vs AG/AA 51.6%). In the lower IPI scores (Low, Low Intermediate) (N=64), lower DFS was found in patients with the AG/AA genotype (835 days, CI 617–1054) vs those with the GG genotype (1483 days, CI 1224–1742) Log Rank p=0.018 (73.5% vs 50%) (Figure 3A).

Figure 3 DFS of DLBCL patients according to genotype of CYP3A5 A6986G in subgroups. (A) IPI score 0–2 (N=64). (B) Cell of origin Germinal Center B-cell (N=49).

On the other hand, the impact of genotypes was analyzed based on the sub-classification of the cell of origin (Han’s algorithm), where GCB cases (N=49) presented lower DFS in patients with the AG/AA genotype (882 days, CI 599–1165), compared to those with the GG genotype (1470 days. CI 1118–1822) Log Rank p=0.040 (65.4% vs 39.1%) (Figure 3B).

The analysis was performed in the high-risk IPI group (intermediate-high, high) and in the non-GCB cell subtype. The results did not show statistical significance.

A Cox regression was performed to analyze risk of presenting and event related to the disease (progression, relapse, death), thus confirming that the GG genotype of CYP3A5 A6986G has been associated to favorable outcomes in terms of maintained complete response after R-CHOP therapy (OR 0.470, CI 0.244–0.903 p=0.023). (Figure 4).

Figure 4 Cox regression analysis between clinicopathological parameters/genotypes and DFS of DLBCL patients (N=102). OR Odds Ratio, LCL Lower Confidence Limit, UCL Upper Confidence Limit. *Significance level p < 0.05.

Discussion

Globally, patients with DLBCL treated with R-CHOP usually have a favorable response rate, exceeding 83%; however, 17% exhibit unfavorable clinical behavior.⁸ In our study, the OS was 83.3%, similar to the 87% reported by Tilly et al in 2022, for patients treated with R-CHOP in a European population. We observed a DFS of 58.8% over a median follow-up time of 855 days, while in Tilly’s study, it was similar (70.2%) over a follow-up of 700 days.⁸

Multiple factors can lead to unfavorable clinical behavior; the most studied by our workgroup are the overexpression of drug resistance genes and the presence of SNPs in genes of the ABC family, as well as drug-metabolizing genes like CYP450.^20,26

In our study population, the most frequent genotype in ABCB1 C3435T was CT, followed by TT and CC (60.78%, 20.59%, and 18.63%, respectively). These results contrast with those reported in Chinese patients with the same pathology, where the most frequent genotype is CT, followed by CC, and the least frequent is TT.^27–29 There are no studies on this polymorphism in DLBCL in which the genotype frequency matches our findings. However, reports in other hematologic neoplasms show similar frequencies, such as Mhaidat et al study in 2011 on Jordanian Hodgkin’s lymphoma (HL) patients.³⁰ In the Mexican onco-hematologic population, Leal-Ugarte et al in 2008 and Olarte et al in 2021, reported similar results to ours: a higher frequency of the CT genotype, followed by TT and, less frequently, CC in pediatric patients with acute lymphoblastic leukemia (ALL) and in adults with acute myeloid leukemia (AML), respectively.^20,31

For ABCG2 C421A SNP, our cohort showed a higher frequency of the CA genotype, followed by CC and AA (51.6%, 36.27%, and 11.76%, respectively). These results are consistent with those reported by Hu et al in 2007 and Kim et al in 2008 in Chinese and Korean DLBCL patients, respectively.^32,33

Regarding the CYP3A5 A6986G polymorphism, our patients showed a higher frequency of the GG genotype, followed by AG, with AA being the least frequent. However, no other reports have addressed this polymorphism in DLBCL patients. In other hematologic malignancies, such as Chronic Myeloid Leukemia (CML), Harivenkatesha et al in 2017 reported a frequency similar to ours.³⁴ In Mexican population without cancer (including conditions like alopecia,³⁵ hypertension,³⁶ respiratory diseases,³⁷ and mental disorders³⁸), various studies have shown similar genotype frequencies for this polymorphism.

Differences in SNP frequencies may be attributed to the geographical location of the studied populations, as well as the methodologies used for detection, such as PCR-ACP, PCR-RFLP, Sanger sequencing, or genotyping by qPCR.

The clinical significance of SNPs in cancer is variable. Several studies have linked genetic variants to a worse prognosis, associated with reduced OS and DFS.^30,39

In hematologic neoplasms, it has been reported that the ABCB1 C3435T SNP in pediatric ALL, multiple myeloma, and lymphoma, the CC genotype is associated with lower OS.^29,40,41 However, in AML patients and breast cancer, the opposite has been reported, the CC genotype was associated with higher OS.^20,42 In lymphoma, our results are consistent with those reported by Hu et al in 2013 and Liu et al in 2020, where no significant association was found between OS and the ABCB1 C3435T SNP.^27,29 The analysis showed no significant association between DFS and this SNP. In contrast, Ni et al 2016, reported that the CC genotype was associated with longer DFS in DLBCL.²⁸ In breast cancer, Li et al in 2017, found that the TT genotype was associated with longer DFS.⁴³

For the ABCG2 C421A SNP, our results showed no significant association with OS or DFS, which is consistent with reports by Hu et al (2007), Kim et al (2008), and Liu et al (2017) in DLBCL.^32,33,44 In contrast, Perrone et al in 2024 found association between the presence of the AA genotype and shorter DFS and OS in DLBCL.⁴⁵ In other cancers, such as prostate and breast cancer, the CC genotype has been associated with better OS and DFS rates, respectively.^39,43 In contrast, Wu et al in 2015, reported higher survival rates in breast cancer patients with the AA genotype.²³ In hematological malignancies such as CML, no association has been found between the genotype and response to imatinib, a substrate of ABCG2.^46,47

Analysis of the ABCB1 C3435T and ABCG2 C421A SNPs did not show any impact on OS, DFS, or clinical characteristics. However, other gene families, such as the cytochrome family and their SNPs, are known to deregulate enzymatic activity, which impacts the metabolism of certain drugs, potentially leading to unfavorable clinical outcomes.⁴⁸ Several studies have shown that the C3435T and C421A polymorphisms in the ABCB1 and ABCG2 genes can alter mRNA expression levels, protein activity, and substrate specificity. A possible explanation for the absence of a significant association between the ABCB1 and ABCG2 SNPs in our study, as reported in other studies, could be the heterogeneity of solid and hematological neoplasms, the diversity of anti-neoplastic drugs administered as part of the treatment, and the genetic variability inherent to each analyzed population and ethnic groups.^18,36,49

Our results showed that the presence of the CYP3A5 A6986G SNP confers longer DFS in patients with DLBCL treated with R-CHOP. In various types of cancer, such as lung and colorectal cancer, the presence of the GG genotype has been described as a favorable prognostic factor for OS and DFS.^50,51 Similarly, in CML, Harivenkatesh et al in 2017, reported a higher rate of treatment failure with imatinib in patients with the AA genotype.³⁴

Odds ratio analysis showed that the presence of the GG genotype of the CYP3A5 A6986G SNP favors a sustained complete response. In DLBCL, there are no reports on the risk associated with this SNP genotype regarding treatment response. However, in hematological neoplasms such as ALL and Acute Myeloid Leukemia/Myelodysplastic Syndromes (AML/MDS), studies by Borst et al in 2011 and Li et al in 2024, reported no clinical implications of this polymorphism.^52,53 Wegman et al in 2007 and Jiang et al in 2016, reported in breast and lung cancer that the presence of the GG genotype is associated with a lower risk of relapse and better treatment responses, similar to our findings.^51,54

The CYP3A5 A6986G SNP has been reported to cause alterations in the metabolism of drugs like cyclophosphamide, vincristine and prednisone (used in the R-CHOP regimen for DLBCL), and therefore impact chemotherapy responses in both solid and hematological cancers.⁵⁵ This could be due to the A>G change in intron 3, which causes a decrease in mRNA expression, leading to a truncated protein with lower catalytic activity, resulting in higher bioavailability of chemotherapeutic drugs and enhancing tumor cell elimination.^14,56,57 In contrast, the AA genotype leads to increased mRNA and protein expression, as well as higher metabolic activity for drug elimination.⁵⁸ Similar findings were reported by Rodríguez-Antona et al in 2007, who found that in T-cell lymphomas, patients with lower expression of other family members, such as CYP3A4, had better survival.⁵⁹

It has been described that, in addition to pharmacological metabolism, CYP3A5 expression in cellular models inhibits cell adhesion capacity, facilitating tumor metastasis and the presence of CTCs in liquid biopsies.^48,60 Other circulating biomarkers, such as circulating tumor DNA (ctDNA), have gained relevance in cancer prognosis; however, genotyping analysis has been shown to have higher sensitivity in CTCs than in ctDNA.⁶¹

The use of liquid biopsies will enable the identification and characterization of CTCs from primary tumors in DLBCL patients, serving as a molecular strategy to detect the genotypes of genes involved in drug resistance and metabolism within the R-CHOP regimen. This approach could facilitate the adjustment of personalized treatments based on these circulating biomarkers. Although no genotyping studies have been conducted on drug-metabolism genes in CTCs, an initial study in lung cancer suggests that this molecular technique can identify EGFR gene alterations in CTCs, which might be utilized not only as a prognostic factor but also to guide treatment decisions.⁶¹

Conclusion

In conclusion, our results demonstrated that the GG genotype of CYP3A5 A6986G SNP, present in CTCs obtained from liquid biopsies of patients with DLBCL, confers longer DFS. Its detection is of great relevance to establish a prognosis and might be implemented as a molecular tool for the follow-up of DLBCL patients. Expanding knowledge of the drug resistance family and cytochrome will allow for the stratification of patients who may not respond to first-line therapy in this type of cancer and other hematologic malignancies.