Centrosomal protein72 rs924607 and vincristine-induced neuropathy in pediatric acute lymphocytic leukemia: meta-analysis

Aim: We examined the utility of the rs924607 TT genotype of the centrosomal protein 72 (CEP72) as a potential biomarker for predilection toward vincristine-induced peripheral neuropathy in children treated for acute lymphoblastic leukemia. Materials & methods: We conducted a random-effects meta-analysis of data from four studies comprising 817 patients. We tested for an association using a recessive model where a one-sided p-value < 0.05 was considered statistically significant. Results & conclusion: We were unable to confirm the association between the rs924607 TT genotype and neurotoxicity (odds ratio: 1.99; p = 0.16; 95% CI: 0.76–5.25) in our global meta-analysis. Analysis of the continuation phase (following induction) studies showed significantly higher odds for neuropathy in CEP72 rs924607 TT homozygotes (odds ratio: 2.28; p = 0.02; 95% CI: 1.16–6.87).

Despite great advances achieved in the last few decades, cancer remains the leading cause of child death by disease in developed countries. Acute lymphoblastic leukemia (ALL) is the predominant childhood malignancy, representing more than 30% of all childhood cancers and approximately 80% of all childhood leukemias. Fortunately, its cure rates have surpassed 85% [1,2]. As more children are cured it is becoming increasingly important to reduce acute and long-term side effects of treatment.
The efficacy of current ALL treatment has been achieved mainly by optimizing the use of existing combinations of chemotherapeutics, rather than adding novel drugs [2]. Therefore, all new information about pharmacokinetics and pharmacogenomics of drugs used in ALL therapy is highly beneficial [3].
Vincristine (VCR) is a very widely used anticancer drug, utilized in pediatric treatment protocols for leukemia as well as solid tumors [2]. It is a vinca alkaloid that causes a mitotic arrest leading to cell death in metaphase. This effect is achieved by disrupting the formation of mitotic spindle microtubules [4]. The major dose-limiting side effect is vincristine-induced peripheral neuropathy (VIPN) which may cause morbidity, necessitate a decrease of VCR dose and thus compromise the effectiveness of treatment [5]. Almost 80% of patients develop VIPN during the treatment and it may affect the quality of their life years after the completion of ALL therapy [6][7][8]. Presently, we do not have reliable biomarkers to identify patients with predilection toward VIPN.
Several studies have assessed various candidate genes involved in VCR metabolism; however, they have not uncovered genetic variants consistently linked with a higher risk of VIPN [9][10][11][12][13][14][15][16][17][18][19][20]. The major focus of our article is a gene called centrosomal protein 72 (CEP72); an overview of other potential biomarkers is given in Table 1. The CEP72 gene polymorphisms as VIPN biomarker A promising VIPN biomarker is a genetic variant in the promoter of the CEP72 gene. A genome-wide association study (GWAS) by Diouf and colleagues established the rs924607 TT as a VIPN risk genotype [9]. It was a prospective study of two cohorts of patients from children's oncology group (COG) and St Jude's Hospital comprising more than 300 children. Homozygous carriers of the CEP72 rs924607 TT genotype had a higher prevalence of VIPN during the 2 years of continuation therapy, which followed induction and consolidation phases (odds ratio [OR]: 2.43; 95% CI: 1.70-3.49) and OR: 4.1; 95% CI: 1.86-9.01 in the St Jude and COG cohorts, respectively. The severity of VIPN was also higher in TT homozygotes [9]. CEP72 encodes a protein crucial for centrosome formation. Using expression studies, Diouf et al. showed that TT homozygotes had significantly lower CEP72 mRNA levels.
They were able to explain this effect by demonstrating that the rs924607 T variant creates a binding site for the NKX-6.3 transcription repressor in the promoter region of the gene. Furthermore, they created pre-B ALL, T-ALL and neuroblastoma cell models and employed short hairpin RNA to impair CEP72 gene expression. These cells were then shown to be more sensitive to VCR [9]. This study was followed by four replication studies, three of which were conducted on pediatric ALL patients and one on adult ALL patients [10][11][12]23]. The three studies on pediatric population produced inconsistent results, with Gutierrez-Camino et al. and Zgheib et al. reporting no association between CEP72 rs924607 and VIPN and Wright et al. reporting a significantly higher risk of VIPN linked with TT genotype (OR: 3.43; 95% CI: 0.93-12.66) [10][11][12]. Such inconsistent results may in part be attributed to distinctive genetic backgrounds of the studied populations and the fact that Gutierrez-Camino et al. limited their study of VIPN to remission induction phase, while the other three studied VIPN across all phases of treatment. Some authors have also highlighted the importance of precise definitions of study end points. Namely, VIPN is most commonly diagnosed and graded using the National Cancer Institute Common Terminology Criteria for Adverse Events which classifies grade 1 as mild, grade 2 as moderate, grade 3 as serious/disabling and grade 4 as life-threatening neurotoxicity. However, these clinical scales are dependent on self-reporting of symptoms which is particularly problematic in children [24]. Alternatively, a recent study by Kavčič and colleagues has shown that VIPN can be objectively measured using electrophysiological studies [25]. Nevertheless, these procedures are uncomfortable for the children and thus unlikely to be used routinely.
A recent GWAS study found no association between the CEP72 gene and VIPN [20]. This study was not included in our meta-analysis because genotyping data was not published.
There is also a promising ongoing clinical trial called Total Therapy XVII for Newly Diagnosed Patients With Acute Lymphoblastic Leukemia and Lymphoma sponsored by St Jude's Research Hospital (ClinicalTrials.gov Identifier: NCT03117751) whose goal is to use strategies based on specific genomic features to improve outcomes for children with ALL and acute lymphoblastic lymphoma [26]. They are using a randomized study approach to determine whether a lower dose of VCR in patients with CEP72 TT genotype (estimated to comprise 16% of the study population) and shorter duration of therapy in patients with CEP72 CC or CT genotype will decrease the incidence and/or severity of VIPN. The researchers are using an unblinded design, where the treating physicians will be aware of the treatment assignments, but the investigators who evaluate neuropathy will not. The study is currently still recruiting patients [26].
In an attempt to elucidate the association between CEP72 rs924607 and VCR toxicity we decided to conduct a meta-analysis of the results of studies on pediatric patients treated for ALL.

Meta-analysis Article selection, inclusion & exclusion criteria
To conduct the meta-analysis of data on the association between VIPN and CEP72 rs924607 genotype TT, we searched the PubMed database for research articles by using the following keywords: 'CEP72 rs924607 and Vincristine', 'CEP72 rs924607 and neuropathy', 'CEP72 rs924607 and Vincristine andneuropathy' as well as 'CEP72 rs924607 and neurotoxicity' and 'CEP72 rs924607 and pediatric cancer'. Subsequently, we searched for cross-references.
Our inclusion criteria were that the study had to report the incidence of VIPN in a pediatric population treated for ALL and also include the CEP72 rs924607 genotyping data. We excluded studies that were done on adult patients and those that were not written in the English language. We did not exclude studies based on the publication date.
From a total of six articles retrieved on PubMed, two were excluded. One was a review article that did not include genotyping data and the other a study on adult patients [3,23]. The final model comprised four studies, including 817 patients, 315 of whom had VIPN.

Data extraction
From the four selected articles we extracted data on the CEP72 rs924607 genotype, incidence of VIPN, VIPN grade, duration of therapy before VIPN occurred and cumulative VCR dose received before VIPN.
Symptoms of VIPN were defined as neuropathic symptoms and sensory, motor and autonomic dysfunction (e.g., impaired tendon reflexes, balance and vibration sensation, altered gait, constipation, orthostatic hypotension). Peripheral neuropathy was graded according to the NCI Common Terminology Criteria for Adverse Events version  1.0. Patients with grades 2, 3 or 4 were considered VIPN cases. Occurrences of VIPN were the end points used for this analysis.

Statistical analysis
To investigate the CEP72 rs924607 VIPN risk genotype TT, we tested for an association between this marker and VIPN using a recessive model. A one-sided p-value < 0.05 was considered statistically significant. We performed a random-effects meta-analysis using the RevMan software [27]. The OR were computed by the Mantel-Haenszel method. The association between the CEP72 rs924607 TT genotype and VIPN was separately estimated in a subgroup of studies that assessed VIPN incidence in the continuation phases of ALL therapy. The continuation phase is defined as a treatment phase following remission induction therapy.

Events description
This study encompasses 315 cases of VIPN from four studies ( Table 2). According to data from prospective and retrospective studies a majority of VIPN cases occurred during the continuation phases of therapy. A majority of VIPN were grade 2 neuropathies (Table 3). Table 4. Results of the random-effects global meta-analysis testing for an association between CEP72 rs924607 TT genotype and vincristine-induced peripheral neuropathy using a recessive model.  ). The graph depicts the odds ratio of each study and its 95% CI (as blocks and lines). The diamond is the total odds ratio with its CI computed by the Mantel-Haenszel method in a random-effects meta-analysis.

Discussion
In 2015, a study by Diouf et al. established CEP72 rs924607 as a promising VIPN biomarker, by reporting a statistically significant association between CEP72 rs924607 TT risk genotype. We analyzed the data from four such studies and revealed the association between the CEP72 rs924607 TT genotype and VIPN in our continuation phases analysis. However, our global analysis did not confirm such association (see Table 4   (generated by RevMan [27]). The graph depicts the odds ratio of each study and its 95% CI (as blocks and lines). The diamond is the total OR with its CI computed by the Mantel-Haenszel method in a random-effects meta-analysis.
However, the failure of Gutierrez-Camino et al. attempt to replicate the discovery study results elucidates another important consideration for further studies. Namely, it is the only study that analyzed only VIPN which occurred during the 4-week induction phase of ALL therapy [10]. Because VIPN predominantly occurs in the later phases of therapy, we decided to perform an additional meta-analysis on continuation phase studies only. In this subgroup our results show a significantly higher risk for VIPN in CEP72 rs924607 TT homozygotes (see Figure 2; n studies = 3; OR: 2.28; p = 0.02; 95% CI: 1.16-6.87).
Wright et al. already published a similar meta-analysis where they reported a statistically significant association between the CEP72 rs924607-TT genotype and VIPN both in their global analysis as well as continuation studies analysis [11]. However, they also included one study on adult ALL patients in their meta-analysis [11,23]. Therefore, the result of our meta-analysis is the first validation of this association exclusively in pediatric patients. Since we were able to replicate this finding in such a diverse pediatric ALL cohort, this pharmacogenomic biomarker could be a valuable tool for clinicians.
Although we believe our study provides valuable insight, there are some drawbacks. Even though we managed to amass a fairly large sample size, we believe including even more patients in future study cohorts would be advantageous. Small sample size and heterogeneous cohorts are common study limitations in pediatric hematooncology and may lead to studies producing contradictory results. A larger sample would make it easier to evaluate the effect of studying children of different ethnic and genetic backgrounds. Furthermore, the majority of the studies included in our analysis were retrospective studies, which is a limitation, as lower grade VIPN may not always be recognized from past medical records, especially in small children. A prospective study where the researchers are actively looking for drug toxicity would have been better suited for the purpose. Therefore, we have high expectations that the Total Therapy XVII for Newly Diagnosed Patients With Acute Lymphoblastic Leukemia and Lymphoma (NCT03117751) clinical trial which is a prospective study and will hopefully further clarify the role of CEP72 in predilection toward VIPN, and its usefulness as a biomarker in clinical practice [26]. We hope to see such studies also on other populations, not just North American, so that we can detect if the findings will be useful in European and Asian settings.
Another important future research direction for the role of CEP72 in VCR metabolism is obtaining confirmatory evidence through functional studies. In the discovery study, Diouf and colleagues attempted to do this by impairing the expression of the gene in pre-B ALL, T-ALL and neuroblastoma cell lines and then subjecting those cells to VCR. They found that these cells were more sensitive to VCR. Furthermore, the CEP72 promoter single nucleotide polymorphism rs924607 genotype T was shown to create a binding site for a transcriptional repressor thus reducing CEP72 expression in human neurons and leukemia cells and increasing their sensitivity to VCR [9]. While this is an interesting finding, we need more studies to understand this relationship.
In the present study, we managed to show that CEP72 rs924607 genotype TT is likely associated with neurotoxicity occurring during the continuation phases (after induction) of pediatric ALL treatment. However, further studies will be needed before we can implement this finding at the bedside and improve patient outcomes in the future.

Future perspective
As high-throughput technologies are providing us with increasingly high-quality data about genomic and transcriptomic attributes of ALL patients and their response to therapy, we are bound to discover even more VIPN biomarkers. In 5-10 years, we hope to have a reliable biomarker for VIPN for the clinical setting, which will allow us to attenuate therapy with VCR in at-risk patients. This will improve both their short-term and long-term quality of life.
CEP72 rs924607 has great potential to become such a biomarker. The Total Therapy XVII for Newly Diagnosed Patients With Acute Lymphoblastic Leukemia and Lymphoma (NCT03117751) clinical trial should provide us with additional answers on this topic. However, in the future, we hope to see more functional studies that will help us understand the mechanisms that cause VIPN and eventually prevent it.

Summary points
• We were unable to confirm the association between the CEP72 rs924607 TT genotype and neurotoxicity in our global meta-analysis. • The analysis of the continuation phase studies exposed a significantly higher risk for neuropathy in CEP72 rs924607 TT homozygotes. • Our meta-analysis is the first validation of the association between the CEP72 rs924607 TT genotype and vincristine-induced peripheral neuropathy exclusively in pediatric patients. • Because we replicated this finding in such a diverse pediatric acute lymphoblastic leukemia cohort, this pharmacogenomic biomarker could be a valuable tool for clinicians. • Developing such biomarkers is a path toward personalized medicine.