Predictive Factors of Oxaliplatin Neurotoxicity: The Involvement of the Oxalate Outcome Pathway

Oxaliplatin, a platinum-based chemotherapeutic agent with a 1,2-diaminocyclohexane (DACH) carrier ligand, displays a characteristic pattern of neurotoxicity, which is the most frequent dose-limiting toxicity, with an acute onset of distal dysesthesia and/or paresthesia, induced or exacerbated by cold and prolonged muscular contraction after a voluntary contraction. Although these symptoms that occur during or short after infusion are transient and generally mild, when treatment is continued, a persistent sensory peripheral neuropathy can develop, eventually causing superficial and deep sensory loss, sensory ataxia, and functional impairment (1). The mechanism of chronic neurotoxicity evoked by some authors is close to that of cisplatin—being the progressive accumulation of a heavy metal, the DACH platin, in neurons and the implication of the detoxication system of glutathione (2).

However, clinical neurophysiologic examinations and nerve biopsy studies have shown that patients displaying sensory symptoms have no signs or very mild signs of axonal degeneration (1). Oxaliplatin has a direct “pharmacologic” effect on the excitability of sensory neurons and muscle cells that has not previously been described with other platinum agents (3). Furthermore, the acute symptoms induced by oxaliplatin are similar to those induced by several drugs or toxins acting on neuronal or muscular ion channels, such as tetrodotoxin (4). Our team and another reported in different cellular models that acute oxaliplatin action on neurons is due to a specific interaction with sodium channels located in the cell membrane (5, 6). Furthermore, we showed first that a peculiar population of Na⁺ channels was involved, probably Ca²⁺-dependent Na⁺ channels, and second that this effect was not reproduced by its cytotoxic metabolite, DACH platinum, but by the other one, oxalate, both the hydrolyzable ligand and a calcium chelator, known for its acute neurotoxic effects, in case of acute poisoning, such as ethylene glycol poisoning (5, 7). Then, oxaliplatin affects voltage-gated sodium channels indirectly through its transformation into one of its metabolites, oxalate, by interfering with intracellular divalent cation regulation in neuronal systems (i.e., calcium and/or magnesium) and by disrupting their intracellular homeostasis (8). Based on these results, we set up an oxalate chelation strategy using calcium-magnesium infusions just before and just after oxaliplatin administration and showed its potential for the prevention of neurotoxicity (9).

In total, two different mechanisms of neurotoxicity have been hypothesized. The classic one is that of heavy metals, such as cisplatin, involving the detoxication process (i.e., the glutathione system), especially glutathione S-transferase π (GSTπ), and certain membrane efflux proteins, such as multidrug-related protein 2 (MRP2; refs. 10, 11). The second mechanism is the repeated and prolonged interference of oxalate with voltage-gated sodium channels, resulting in acute or chronic neurotoxicity (12).

Oxalate synthesis and outcome are closely linked to those of glyoxylate, itself produced from serine, glycine, and alanine metabolisms, used by the liver to produce glucose (13, 14). Glyoxylate is detoxified by alanine glyoxylate transferase (AGT), exclusively in peroxisomes, through its transformation in glycine in a reaction coupled to alanine conversion in pyruvate. In the cytosol, glyoxylate is metabolized into glycolate by glyoxylate reductase-hydroxypyruvate reductase (GRHPR; Fig. 1). The main role of this enzyme is to prevent the accumulation of glyoxylate in the cytosol, converting it into glycolate (15). The last catabolic pathways of glyoxylate are its transformation in oxalate by lactate dehydrogenase in cytosol. It usually remains minor because AGT detoxifies most of glyoxylate, consequently considerably limiting its oxidation in oxalate.

A deficiency in one of these two enzymes, AGT and GRHPR, known as genetic disease, can lead to an abnormal accumulation of oxalate called hyperoxaluria (16). AGXT gene variants are very rare, but besides the genetic disease, two variants are frequent, the wild-type known as a major allele and less frequent minor allele, found in 20% of Caucasian patients, a haplotype characterized by three mutations: two substitutions Pro11Leu and Ile1142Met and a duplication of 74 bp (17). Only Pro11Leu single nucleotide polymorphism has a phenotypic effect. It reduces AGT catalytic activity by 3 in homozygote patients (i.e., 4% of Caucasian patients; ref. 18).

GRHPR variants, leading to the formation of a truncated protein or to the loss of catalytic enzyme activity (965T>G, 103 delG), are very rare (19).

Because glyoxylate is an intermediate step of the two metabolic pathways of oxalate, we hypothesized that patients with a partial deficiency of one of these enzymes could be at high risk of neurotoxicity after oxaliplatin infusion because they would not be able to manage repeated acute and high levels of oxalate production.

In the present study, we wanted to further characterize the oxalate outcome and effect after oxaliplatin infusion and to determine predictive factors of neurotoxicity in clinical practice that could allow their pretherapy detection and leading to neuroprevention.

Therefore, we carried out two prospective studies. First, we sought to characterize the metabolic effect on patients treated with oxaliplatin. We tested urine for oxalate, cations, and amino acids involved in the oxalate and the glyoxylate metabolism. Second, we looked for a correlation between oxaliplatin-induced neurotoxicity and polymorphisms of the two most important enzymes, AGT and GRHPR, involved in the metabolic pathway of oxalate, the glyoxylate-oxalate metabolism. On the other hand, we investigated in the same population of patients the hypothesis of a cisplatin-like neurotoxicity genesis, involving a deficiency in the cell detoxification pathway of heavy metals, especially the glutathione cycle and the ATP-binding cassette family. We characterized polymorphisms of GSTπ and MRP2, also called ABCC2 (10, 11, 20).

These two observational prospective studies were carried out on French Caucasian patients treated for advanced carcinomas with biweekly oxaliplatin + 5-fluorouracil and folinic acid (FOLFOX regimen; ref. 21): 85 mg² oxaliplatin on day 1 plus a 5-fluorouracil 400 mg/m² bolus and 2,500 mg/m² 46-h infusion plus 200 mg/m² folinic acid on days 1 and 2. In the second study, with FOLFOX regimen, taking into account the interest of calcium and magnesium infusions in term of neurotoxicity prevention, patients received 1 g of calcium gluconate and 1 g of magnesium sulfate in 15-min infusion just before and just after oxaliplatin infusion, as published previously (9). To be eligible, patients had to be naive of oxaliplatin, to have a WHO performance status <2, no preexisting neuropathy, a life expectancy of at least 3 months, an age lower than 80 years, and adequate hematologic and cardiac status.

The relevant Ethical Committee stated that this observational study was not submitted to the French Huriet-Serusclat law. Written informed consent was obtained from all patients before pretherapeutic genotyping (135 patients) or urine sampling (10 patients). Results of pretherapeutic genotyping were not taken into account. The 5-fluorouracil dose was adjusted individually, as is usually done in our institution as routine practice, beginning with the second cycle based on a pharmacokinetic follow up, as previously published (22). Fluorouracil concentrations in plasma were measured in the Oncopharmacology Department (23).

Patients were examined biweekly, with historical and physical examination. They were asked for treatment tolerance and underwent examinations with hemograms, ionograms, and liver and kidney tests before every chemotherapy cycle.

Response evaluation was carried out after three cycles by comparing tumor measurements before and after 3 months of treatment. Treatment was then prolonged up to a total of at least 6 months, except for progressive disease.

All adverse events were taken in account and graded for severity according to the National Cancer Institute Common Toxicity Criteria (NCI-CTC) scale. Treatment had to be stopped in case of grade 4 toxicity and the French Drug Committee had to be notified. The safety assessment of oxaliplatin + 5-fluorouracil/leucovorin treatment included the following. (a) Neurologic safety: Acute and chronic, peripheral and perioral neuropathy, pseudolaryngospasm, laryngeal dysesthesia, muscular contractions, motor troubles, stiffness, and muscular pain. Both the NCI-CTC version 1 scale and the specific Levi-modified neurotoxicity scale were used (24). Actually, although the Levi-modified scale seems very useful for treatment adjustment and neurotoxicity managing, NCI-CTC scale remains more suitable for neurotoxicity assessment. (b) Other toxic side effects, such as myelotoxicity and digestive tract toxicity, especially diarrhea. (c) Other general variables, such as asthenia and weight loss.

The cumulative doses of oxaliplatin, dropouts, for whatever reason were also assessed. In the event of a case of significant grade 2 neurotoxicity, the oxaliplatin dose remained unchanged, but the cycle could be delayed until resolution. In cases of recurrence, the dose was reduced by 20%. In the case of grade 3 toxicity, treatment was interrupted until resolution of the toxic manifestations and then restarted with a 20% decrease of the dose. In the event of a significant grade 2 hematotoxicity, the oxaliplatin dose remained unchanged, but the cycle was delayed until resolution. In case of recurrence, the dose was reduced by 20%. In the case of grade 3 toxicity, treatment was interrupted until resolution of the toxic manifestations and then restarted with a 20% decrease of the dose.

First study. We measured (a) Ca²⁺, Mg²⁺, Na⁺, and K⁺ concentrations in plasma at H0, H2 (before and just after the end of oxaliplatin infusion), H5, day 8, and day 15 and (b) oxalate, Ca²⁺, Mg²⁺, Na⁺, and K⁺ concentrations in urine, weighed by urine creatinine at H0 (before oxaliplatin infusion), H5, day 8, and day 15. Urine samples were collected as follows: urine samples of the previous 24 h before oxaliplatin administration for H0, urine samples between H0 and H5 for H5, and 24-h urine samples for days 8 and 15.

We also characterized the urinary elimination of different amino acids with urine amino acid chromatography.

Second study. We genotyped AGXT, GRHPR, GSTπ, and MRP2. We looked for the most relevant mutations of the AGXT gene. They are listed on Table 1. Some of them are rare, involved in type 1 hyperoxaluria (15, 16). They are located on exons 1, 4, 7, and 10. Three of them are more frequent and characterize two haplotypes: the “major haplotype” and the less frequent “minor haplotype” (25). They are constituted of three variants located on exon 1, 154C>T (Pro11Leu), on exon 4, the duplication of 74 bp, and on exon 10, 1142A>G (Ile340Met). The minor haplotype is found in 20% of northern American and European populations (25).

On the GRHPR gene, we looked for six variants leading to splicing errors and to a truncated protein with the loss of cofactor or enzyme substrate binding site or the loss of catalytic activity: four variants located on exons 2, 6, and 9 and two variants on introns 4 and 7: 103delG located on exon 2, del AAGT on intron 4, and four single nucleotide polymorphisms: 295C>T (Arg99Stop), 965T>G (322Met>Arg), 494G>A (165Gly>Asp), and G>A on the 3′ splicing region of intron 7 (26).

About GSTπ, two variants were looked for: 313A>G (exon 5, 105Ile>Val) and 341C>T (exon 6, 114Ala>Val). Four haplotypes have been defined: GSTP1A (Ile105-Ala114), GSTP1B (Val105-Ala114), GSTP1C (Val105-Val114), and GSTP1D (Ile105-Val114; refs. 10, 11). As results in the literature about their implication in the pathogeny of neurotoxicity were contradictory, we looked for these GST variants.

For MRP2, we looked for two variants: 224C>T and 3972C>T (20).

Blood samples were obtained for DNA isolation and the determination of genotypes. All procedures were reviewed and approved by accredited ethics review boards, and patients signed informed consent form. DNA was extracted from peripheral blood mononuclear cells (500 μL of whole blood) using a DNA isolation kit for blood/bone marrow/tissue (Roche Molecular Diagnostics). Each sample was controlled with respect to DNA isolation by UV transillumination of ethidium bromide–stained gels from subsequent electrophoretic separation in 1.2% agarose.

Pyrosequencing analysis

PCR conditions. PCR was done with an initial denaturation for 5 min at 95°C followed by 50 cycles of denaturation for 30 s at 95°C, primer annealing for 30 s at 60°C, and extension for 1 min at 72°C followed by a final extension for 5 min at 72°C. All amplification reactions were done in a DNA Thermal Cycler Eppendorf, with 1 unit of Taq polymerase (Euroblue Taq, Eurobio).

PCR conditions were the same for the different mutations tested. Annex 1 (Table 5) shows the different sets of primers used to amplify the sequences of interest.

Choice of the sequencing primers. Different sequencing primers were designed to carry out AGXT, GRHPR, GSTπ, and MRP2 gene pyrosequencing analysis. Then, a selection was made based on the ability to provide interpretable Pyrograms. DNA products consisted of amplified genomic DNA from control subjects.

Conditions for the pyrosequencing analysis. Templates for the pyrosequencing analysis were prepared as recommended by the manufacturer. Real-time pyrosequencing was done at 28°C in an automated 96-well pyrosequencer using PSQ SNP 96 enzymes and substrates (Pyrosequencing AB). Before analysis, the enzymes and each of the four deoxynucleotide triphosphates (PSQ 96 SNP Reagent kit, Pyrosequencing AB) were loaded into a special cartridge that was mounted in the PSQ instrument. The 96-well microplate on which samples were annealed with their respective sequencing primers is designed to fit onto a socket in the PSQ 96 system. To apply the pyrosequencing technique to our work and to provide accurate results, different variables were modified and tested (27, 28).

Frequencies of the different gene variants and the toxic side effects were measured and then compared. Correlations were looked for to assess whether statistical difference existed between two observations (i.e., between two percentages relative to toxic event occurrence). Pearson's χ² test was used.

Neurotoxicity, the outcome variable, had four categories: grades 0, 1, 2, and 3. This ordinal variable was crossed with each categorical variable and tests of χ² allowed assessing statistically significant associations. Taking into account the ordinal nature of the outcome variable, Kruskal-Wallis or Mann-Whitney tests were done to complete the χ² tests.

The age, the cumulative administered dose, and the body surface were assumed quantitative, and then their different associations with the neurotoxicity were examined using one-way ANOVA. For a comparison between two groups (e.g., grade 0-1 versus grade 2-3), the t test was used.

Stepwise multivariate logistic regression model was used to assess the ability of covariables to predict neurotoxicity. The χ² goodness of fit, odds ratios, and 95% confidence intervals were provided. In this analysis, covariables were the statistically significant (P < 0.05) associated variables in the univariate analyses.

By using the covariables, this method allowed an optimal discrimination between the patients who had neurotoxicity (grades 2 and 3) and those who had not (grades 0 and 1). At each step, the covariable that brought most of information got in the model. Then, the diffuseness of information made by two associated covariables was avoided. Age and cumulative dose administered were recoded in categorical variables in this analysis.

Moreover, to evaluate each pretherapeutic variable as a predictive factor of toxic side effects, we determined the sensitivity, specificity, and positive and negative predictive values of variants. First, to check that the tests ordered the patients in true positive, true negative, wrong positive, and wrong negative, we crossed the tests and calculated the κ concordance coefficient.

Sensitivity was calculated as follows: number of true positive patients/number of true positive + wrong negative patients.

Specificity was calculated as follows: number of true-negative patients/number of true negative + false-positive patients.

Statistical Package for the Social Sciences 13.0 and 10.0 for Windows (SPSS, Inc.) were used for statistical analyses.

First study. Ten initial patients treated for metastatic colorectal cancer were studied. Their mean age was 60 ± 7 years. They were followed for 3.7 cycles (1-6 cycles).

As shown on Table 2, we found no difference in cation concentrations in plasma after oxaliplatin administration compared with baseline levels. We did not do oxalate measurement in plasma. In urine, we observed a significant increase of oxalate excretion within 5 h following oxaliplatin infusion, way over baseline levels (Table 2). These returned to the baseline levels on days 8 and 15. Table 2 also shows an increase of cation excretions (calcium, sodium, and potassium) 5 h after oxaliplatin administration followed by a decrease on day 8 for some patients and a return to the baseline level on day 15.

In fact, oxalate concentrations in urine did not display the same evolution for these 10 patients. For 3 of them, it remained stable, whereas for the others it increased up to a factor of 10. The urine excretion of calcium and magnesium closely followed that of oxalate.

The infusion of sodium chloride with oxaliplatin does not explain the elevation of cation levels in urine.

Chromatograms of the amino acids in urine were done before and after oxaliplatin infusion. Table 2 shows the concentrations of 4 amino acids out of the 20 main ones, which displayed modifications of concentrations in urine after oxaliplatin infusion: glycine, alanine, serine, and taurine. Interestingly, they are linked to the metabolism of glyoxylate, especially via the AGT enzyme. We observed a significant augmentation of their concentrations in urine, but for two patients, they remained stable, whereas for the eight other patients the increase could be up to 4-fold. This increase was accompanied by the simultaneous increase of urinary excretion of oxalate, calcium, and magnesium. However, we investigated too few patients to look for a correlation with neurologic symptoms.

Second study. We investigated a total of 135 patients in a prospective study: 93 men and 42 women. Their characteristics are displayed in Table 3.

Toxicity evaluation. Seventeen patients experienced severe neutropenia, 13 grade 3, and 4 grade 4. Ten patients had grade 3 diarrhea. One patient presented an allergic shock at the beginning of the fifth oxaliplatin infusion.

Neurologic toxicity has been especially characterized, evaluated, and graded, with its two types, acute and chronic, being identified.

Eighty-four patients experienced acute neurotoxic effects within NCI-CTC grades and 25 of them experienced grade 2 or 3 neurotoxicity. Twenty-six patients suffered from chronic neuropathy and 24 of them had grade 2 or 3 neurotoxicity (Table 3). If we look more precisely at the grade of neurotoxicity, 19 patients had grade 2 and 6 had grade 3 acute toxicity compared with 13 and 11 who had grade 2 and 3 chronic neurotoxicity, respectively (Table 3). Sixty-eight patients presented only distal neurotoxic effects, whereas 9 had also dysesthesia and/or paresthesia of the perioral and oral region, 7 had pharyngolaryngeal tract dysesthesia, and 6 had trismus for several days.

Genotyping results. About the GRHPR gene, no variant was found (Table 4). In the AGXT gene, we only found the common polymorphism (Table 4) composed of the three variants that constitute the haplotype “minor allele”: the 154C>T substitution, the duplication of 74 bp located on exon 1, and the 1142A>G substitution on the exon 10. In our population of patients, almost all the patients with the variant 154C>T, except three, presented both 74-bp duplication and 1142A>G (Table 4, bottom).

The frequency of their combination was 31.5%, which is higher than the 20% previously reported in the literature in a Caucasian population (Table 4; refs. 18, 19).

Usually, the minor allele is considered as present if at least two variants are found out of the three. We found 26.8% heterozygotes and 4.7% homozygotes; this agrees with previous results (∼4% homozygotes).

We investigated GSTπ polymorphism too. We found 64 GST1A patients, 14 GST1B patients, 1 GST1C patient, no GST1D patient, and 71 heterozygote patients.

About MRP2, we looked for MRP2 24C>T and 3972C>T variants. Seventy-four, 51, and 10 patients were wild-type homozygotes, heterozygotes, and variant homozygotes, respectively, for 24C>T polymorphism. Forty-seven, 58, and 30 patients were wild-type homozygotes, heterozygotes, and variant homozygotes, respectively, for 3972C>T polymorphism.

Relationship between demographic characteristics, genetic polymorphism, and neurotoxicity. We looked for clinical and biological predictive factors of neurotoxicity. The influence of the main demographic characteristics on the variable of interest, neurotoxicity, was examined. There was neither significant association between age (P = 0.059, one-way ANOVA) nor between gender (P = 0.404, Fisher's exact test; grades 0 and 1 versus grades 2 and 3) and neurotoxicity (Table 5A). The other factor of interest in study was the administered dose. We compared the means of cumulative dose administered in different groups of neurotoxicity (grades 0, 1, 2, and 4) and found statistically significant differences (P = 0.001, one-way ANOVA; Table 5A).

The pharmacogenetic approach consisted in seeking two polymorphisms of the genes of the main enzymes responsible of oxalate catabolism. We looked for a relationship between AGXT, GSTπ, and MRP2 genotype and neurotoxicity. AGXT relevant variants were the 154C>T and 1142A>G genotypes. In this population, the frequencies of the C/C allele, C/T allele, and T/T allele were 69.4%, 25.8%, and 4.8%, respectively. There were 69.9% with the wild-type (A/A), 25.3% of heterozygous (A/G), and 4.8% of homozygous (G/G) variant genotypes. It was hypothesized that the presence of mutations on these genes increased the risk of oxaliplatin-induced toxicity. About AGXT, we found that patients with the minor allele AGXT haplotype had a significantly higher risk of acute neurotoxicity. Only 4 patients of 92 (4.3%) who presented the major haplotype, the wild-type, had acute neurotoxicity at grade 2 or 3 compared with 23 of the 43 (53.4%) patients who had the minor haplotype.

On the other hand, 87.2% of subjects who had not any neurotoxicity (grade 0) were from wild-type (C/C) group; this percentage was 10.6% for patients with (C/T) genotype and 2.1% for those with the (T/T) variant genotype. In contrast, no patient was wild-type out of those who had severe neurotoxicity (grade 3), but 83.3% in this group had the heterozygous (C/T) variant genotype and they were 16.7% with the homozygous (T/T).

Interestingly, of seven homozygote variant patients, four presented severe neurotoxicity. Of the three other ones, two had initial reduced oxaliplatin doses and the third one had to have his treatment stopped after 2 months because of a grade 4 leucopenia.

These results concluded to very statistically significant relationships between neurotoxicity and 1142A>G and 154C>T genotypes (χ² = 39.33, P < 0.001, C/C versus C/T + T/T; χ² = 38.88, P < 0.001, A/A versus A/G +G/G; Table 5B).

We found equivalent results with 1142A>G genotype (Table 5B). Sensitivity, specificity, and positive and negative predictive values of the detection of the minor haplotype as predictive factors for acute neurotoxicity were 0.83, 0.84, 0.53, and 0.95, respectively.

Because both genotypes 1142A>G and 154C>T were associated to neurotoxicity, multivariable logistic regression analysis was done to report the specific information of each covariable on the outcome variable, neurotoxicity. The covariables included in the final model were the 154C>T genotype and the cumulative dose. The results showed that 60.7% of the variations were explained and 91% individuals were correctly classed by the model.

It is noteworthy that we found an identical relationship between minor haplotype and chronic neurotoxicity. Only 5 patients of 92 with the major haplotype experienced a chronic neurotoxicity compared with 23 of 43 with the minor haplotype. Again, the difference was highly significant.

We looked at GSTπ and MRP2 polymorphisms and found no correlation between their variants and neurotoxicity, either acute or chronic (P < 0.410; Table 5B).

According to the objectives of the study, we were able to confirm the biological importance of oxalate production from oxaliplatin, and to determine relevant predictive factors of neurotoxicity, based on the exploration of the oxalate outcome pathway.

Indeed, oxaliplatin has a unique profile of induced neurotoxicity, with two distinct types, acute and chronic, whose relationship remains unclear to date. These very peculiar characteristics of the acute neurotoxic effects of oxaliplatin previously led us to suspect the implication of ionic channels. Both our team and another one reported with different models its functional effects on sodium channel functions (6, 7). Furthermore, we showed with an electrophysiologic approach that one of its two metabolites, the oxalate, a calcium chelator, was clearly involved and had a deleterious effect on specific sodium neuron channels, voltage gated, and calcium dependent (6). We showed at the same time that DACH platin, the other oxaliplatin metabolite, had no effect on the ionic channels (6). Additionally, according to our hypothesis on oxalate involvement, we reported the interest of calcium gluconate and magnesium sulfate infusions to prevent oxaliplatin neurotoxicity in both its acute and chronic manifestations (9). Calcium and magnesium significantly reduced the incidence of neurotoxic effects, principally that of the acute ones.

Besides, some authors hypothesized a neurotoxic mechanism similar to that of cisplatin due to the accumulation of DACH platin in neurons and its effect on their trophicity (29). This hypothesis has never been confirmed. Actually, little is known about the molecular mechanisms of cisplatin neurotoxicity itself. Mollman (30) suggested that axonal transport may be involved, as the neurotoxicity progresses for weeks after treatment is stopped. More recently, it has been suggested that cisplatin may damage dorsal root ganglion neurons, containing the cell bodies of sensory neurons, shown to be the most vulnerable structure, by inducing apoptosis, based on in vitro and in vivo experiments in the rat (31, 32). Although it is hardly surprising that neurotoxic injury to the nervous system ultimately produces neuronal degeneration and cell death, the triggering mechanisms remain to be elucidated.

These two hypotheses of neurotoxicity generated two kinds of approaches for treating or preventing neurotoxicity. Certain authors tested unsuccessfully glutathione, amifostine, or α-lipoic acid, increasing the pathway of heavy metal detoxication (33–35). The last attempt with xaliproden did not provide convincing results (36). The other approach was to block the action of oxaliplatin on sodium channels with carbamazepine or gabapentin or to chelate oxalate by i.v. infusions of calcium and magnesium, the latter now being widely used in clinical practice (9, 37, 38).

In the present study, we further investigated the “track” of oxalate. We showed that its concentrations in urine could be considerably increased, up to 10-fold, after oxaliplatin administration and that it was accompanied by elevated calcium, magnesium, and sodium urinary excretions. This was also coupled to modifications of the urinary excretion of four amino acids: glycine, alanine, serine, and taurine. The analysis of the amino acid chromatograms in urine suggested an interaction between oxalate and amino acids, with a global profile showing a simultaneous increase, but at variable degrees, of their concentrations. All of them are closely linked to each other as well as with oxalate and glyoxylate metabolisms: glycine comes from glyoxylate by AGT; alanine and serine, as amine group providers, are both substrates and products of AGT; and taurine is closely linked to serine (39).

These results confirm that, after oxaliplatin administration, the liberation of oxalate is not negligible because it can be measured in urine and provokes a cascade of enzyme reactions of the glyoxylate metabolism and also modifications of urinary excretions of certain amino acids linked to it.

Therefore, our hypothesis is that, after a quick oxaliplatin infusion, the important and brutal afflux of oxalate would result, due to the mass action law, in the production of glyoxylate by lactate dehydrogenase. AGT would induce glycine formation and consequently the transfer of the amine group. This could explain the concomitant variations of alanine, serine, and taurine.

We observed wide individual variability in the management of the rapid liberation of oxalate because after oxaliplatin the range of increase of oxalate in urine was considerable from 1- to 10-fold. The profile of calcium and magnesium in urine was closely linked as well as those of the four amino acids involved in the oxalate-glyoxylate metabolism. These results suggest that some patients were able to quickly metabolize oxalate through AGT and GRHPR, whereas other patients could not and consequently had a major increase of urinary excretion of oxalate.

We therefore looked for polymorphisms of the genes coding for the two major enzymes involved in oxalate outcome, AGT and GRHPR, and their eventual effect on oxalate management. We used the pyrosequencing method for its reliability and accuracy and because it is suitable for clinical practice, requiring 500 μL of total blood to extract DNA for genotyping (40).

In the population of 135 patients, thanks to the calcium and magnesium infusions, acute neurotoxicity frequency was lowered: 20 patients experienced grade 2 and 7 patients had grade 3 neurotoxicity. Chronic neurotoxicity was also reduced but at a lesser extent. About the genotyping study, we found no variant of the GRHPR gene in our population of 135 patients but we did find major and minor haplotypes of AGXT: the 154C>T substitution, the duplication of 74 bp located on exon 1, and the 1142A>G substitution on the exon 10 (15, 17).

Almost 32% of our population of patients presented the minor haplotype, with 4.7% being homozygotes (17, 18). According to the literature, out of the minor haplotype, the variant 154C>T (Pro11Leu) located on exon 1 plays a major role in the reduction of AGT activity, being ∼30% less than its baseline due to a location error of the enzyme in the mitochondria rather than in the peroxisomes (18).

Patients with the minor allele AGXT haplotype were at significantly higher risk of acute neurotoxicity, 53.4% versus 4.3%, despite calcium and magnesium infusions. Moreover, of seven homozygote variant patients, four presented severe neurotoxicity, and of the three others, two had reduced cumulative dose of oxaliplatin. Interestingly, we found that minor haplotype cases were similarly linked to chronic neurotoxicity: 52% versus 5.5%. This result clearly favors a common mechanism in the genesis of acute and chronic neurotoxicity.

Besides the involvement of oxalate in the genesis of neurotoxicity, we explored the alternative hypothesis of a cumulative cisplatin-like toxicity, implying a detoxifying process, such as GSTπ or MRP2. We investigated their polymorphisms, already reported by different authors, with contradictory results (10, 11, 20). About GSTπ, we found frequencies close to those reported in the literature, but we found no correlation between any haplotype and neurotoxicity, neither acute nor chronic. Our results differ from those of Lecomte et al. and Grothey et al. who reported in a smaller population a correlation between the two variants and the incidence of chronic and acute neurotoxicity, respectively (10, 11). However, previous results are not in favor of a cisplatin-like mechanism of toxicity. Unlike cisplatin, sural nerve biopsy studies on patients having received few cycles of oxaliplatin and displaying neuropathic symptoms had shown no signs or only minor morphologic signs of axonal degeneration, which could hardly be distinguished from normal findings and which contrasted with the intensity of subjective symptoms. Thus, oxaliplatin is likely to have a direct pharmacologic effect on the excitability of sensory neurons and possibly muscle cells that has not been previously described with other platinum agents. Besides, certain authors showed that DACH platin, the other metabolite of oxaliplatin, was not involved in neurotoxicity (41), and Takimoto et al. (42) reported that oxaliplatin neurotoxicity was neither correlated to creatinine nor to ultrafilterable platinum clearances and was not linked to an accumulation of platinum. Thus, functional neuronal damage produced by oxaliplatin may result in part from the effects of this drug on voltage-gated sodium channels and chronic oxaliplatin-induced neuropathy could be the long-term consequence of its acute toxicity. On the other hand, cisplatin-induced neuropathy is classically reported as a very long-term platinum retention in deep compartments, such as neuronal tissue, and to progressive accumulation (41), whereas oxaliplatin has a completely different pharmacokinetic profile and does not accumulate in plasma, with repeated chemotherapy cycles (43).

About MRP2, previous data were very scarce. We looked for two single nucleotide polymorphisms, 24C>T and 3972C>T, and found that neurotoxicity, both acute and chronic types, was equally distributed in the different variants.

In total, it is likely that, in patients who displayed the AGXT minor haplotype, the major enzyme of glyoxylate and oxalate metabolism has a partially reduced activity. These patients seem unable to cope with a brutal inflow of oxalate after oxaliplatin infusion, whose elimination becomes predominantly urinary. Its intracellular concentration increases and it interferes with sodium channels and generates acute neurotoxicity.

The disruption of ionic movements could hinder processes such as gene expression and lead to heterogenous mRNA synthesis, intraneuronal heterogeneity of the populations of Na⁺ channels in cutaneous afferents, membrane instability, and possibly ectopic impulse generation, resulting in paresthesia, dysesthesia, and hyperesthesia (12). With repeated oxaliplatin infusions, oxalate could progressively affect neurotransmitter release, growth cone elongation, and neuron trophicity and thus could lead to chronic neurotoxicity. Indeed, this favors a common pathogeny of both acute and chronic neurotoxicity due to oxaliplatin.

These data support the future use of this genotyping approach, suitable to clinical practice, to be used as a screening test to identify patients at high risk of neurotoxicity before oxaliplatin-based therapy. This could help better prevent neuropathy by, for example, being more prudent in terms of cumulative oxaliplatin dosage in the case of a patient with the minor haplotype or by using higher doses or more prolonged calcium-magnesium administrations rather than carbamazepine, gabapentin, or venlafaxine, whose efficacy is not obvious and whose toxicity is a cause of concern.