Determinants of the interindividual variability in serum cytidine deaminase activity of patients with solid tumours

Abstract

Aims

Cytidine deaminase (CDA) activity in cancer patients' serum has been proposed as a predictive biomarker for efficacy and toxicity of nucleoside analogues. However, discrepant results about its predictive value have been reported due to the high interindividual variability in CDA activity. This study aimed at identifying determinants of this interindividual variability.

Methods

From December 2014 to November 2015, 183 patients were prospectively included. Serum CDA activity, biological and clinical characteristics as well as five common single nucleotide polymorphisms (SNPs) in the CDA gene (c.‐451C > T, c.‐92A > G, c.‐33_‐31delC, c.79A > C, c.435 T > C) were analysed. Associations between clinical characteristics, pharmacogenetic variants and CDA activity were univariately tested. P < 0.1‐candidate variables were analysed through a multivariate analysis. The association between CDA activity and toxicity was assessed for the 56 gemcitabine‐treated patients. Intraindividual variability in CDA activity was explored in six pancreatic cancer patients treated with gemcitabine.

Results

Median CDA activity was 3.97 U mg^–1 (range 1.53–15.49 U mg^–1). A univariate analysis showed that CDA activity was statistically associated with Eastern Cooperative Oncology Group performance status, mild or severe malnutrition, inflammatory syndrome, leucocyte count, neutrophil count, albumin, C‐reactive protein and ‐c.‐33_‐31delC single nucleotide polymorphism. A multivariate analysis identified that only neutrophil count (P < 0.0001) and severe malnutrition (P = 0.0278) were independently associated with CDA activity. Low CDA activity (<2 U mg^–1) was not statistically associated with severe gemcitabine‐related toxicities (P = 0.16). A decrease in CDA activity was observed during the longitudinal follow‐up of six pancreatic cancer patients treated with gemcitabine (P = 0.03).

Conclusions

These results suggest that neutrophil count and malnutrition should be considered for the interpretation of pretherapeutic CDA activity.

What is Already Known about this Subject

• Serum cytidine deaminase (CDA) activity in cancer patients is a potential biomarker for the efficacy and toxicity of nucleoside analogues.

• Discrepant results have been reported due to the high interindividual variability of CDA activity.

What this Study Adds

• This study shows for the first time that elevated neutrophil count and malnutrition are significantly correlated with increased serum CDA activity in cancer patients.

• Our work suggests that serum CDA activity should be interpreted along with the inflammation and malnutrition status.

Introduction

Human cytidine deaminase (CDA) is a cytoplasmic enzyme involved in pyrimidine nucleotide synthesis. It is mainly synthesized in the liver and the placenta but a high concentration of CDA is also found in mature neutrophils 1, 2. Its physiological role is to maintain the pyrimidine nucleotide cell pool by catalysing the deamination of cytidine or deoxycytidine into uridine or deoxyuridine, respectively 3. CDA is also involved in the metabolism of nucleoside analogues such as gemcitabine, capecitabine, cytarabine and azacytidine, which are widely used to treat solid and haematological neoplasms. By catalysing the deamination of nucleoside analogues, CDA induces their deactivation (gemcitabine, cytarabine, azacytidine) or their activation (capecitabine) 4, 5, 6. Antileukaemia activity of cytarabine has been associated with CDA activity within the leukaemic blasts 7, 8. In patients with solid tumours and in a mice model, serum CDA activity was inversely correlated with gemcitabine exposure 9, 10. Interestingly, patients with very low serum CDA were consistently reported to experience severe toxicities of gemcitabine 10, 11, 12, while severe toxicities of capecitabine were observed in patients with high CDA activity. Serum CDA activity could also influence the clinical activity of nucleoside analogues since patients with the higher range of CDA activity and treated with a gemcitabine‐, 5‐azacytidine‐ and decitabine‐based chemotherapy had a worse outcome 7, 8, 13, 14, 15, 16. These results suggest that serum CDA activity could be considered as a circulating biomarker for developing adaptive dosing strategies with nucleoside analogues 17.

A large interindividual variability in serum CDA activity has been reported 10, 13. The determination of CDA gene polymorphisms has shown promising but discrepant results with regards to their correlation with serum CDA activity and clinical activity of nucleoside analogues 9, 10, 12, 13, 14, 16, 18, 19, 20, 21, 22, 23, 24, 25. This discrepancy between serum enzyme activity and CDA gene polymorphisms suggests that other factors are likely to contribute to the large interindividual variability in serum CDA activity. Unfortunately, data about the identification of these factors are sparse. A large part of the interindividual variability in serum CDA activity remains unexplained 16, 26, 27, even though it might be a potential source of unpredictable toxicities of nucleoside analogues. Thus, identification of these factors may help tailoring nucleoside analogue‐based therapies.

The present study aimed to identify clinical, biological and genetic parameters contributing to the interindividual variability in CDA activity in serum from a large cohort of chemotherapy‐naïve patients with advanced solid tumours, whatever the anticancer drug they were to receive as first‐line treatment. Secondly, this study aimed to: (i) investigate the relationship between serum CDA activity and occurrence of grade ≥3 adverse events in cancer patients treated with gemcitabine; and (ii) monitor serum CDA activity in pancreatic cancer patients treated with gemcitabine monotherapy during the first 3 months of treatment.

Methods

Patients

From December 2014 to November 2015, unselected consecutive patients with advanced solid tumours were prospectively included in a study cohort, regardless of tumour location and chemotherapy regimen. CDA activity was assessed before the initiation of chemotherapy. This prospective study conducted in an outpatient chemotherapy unit at Cochin hospital (Paris, France) was approved by the local Review Board for Oncology. All patients provided written informed consent, in compliance with the ethical principle of the revised Declaration of Helsinki (Edinburgh, 2000) and according to French regulations.

Demographic, biological and clinical data were collected at the time of blood collection for CDA activity assessment: sex, age, body mass index, active smoking, Eastern Cooperative Oncology Group performance status, malnutrition, inflammatory syndrome, liver disease, metastatic disease, liver metastasis, primary tumour location, haematological parameters, liver enzymes, total bilirubin, albumin, C‐reactive protein (CRP), ferritin, creatinine and planned chemotherapy. Malnutrition was defined according to the guidelines of the French Society of Clinical Nutrition and Metabolism (Supplemental Table S1) 28. Inflammatory syndrome was defined as increased levels of CRP, absolute neutrophil count or ferritin.

CDA phenotyping analysis

A 5‐ml blood sample was collected in a dry tube at any time in the day. Clotted blood samples were centrifuged at 1850× g for 5 min (4°C). After serum collection, total serum protein level was assayed on Cobas 8000 (Roche Diagnostics, Meylan, France) before storage at −80°C until CDA activity assay. CDA activity in serum was assayed as previously described 29. Serum from patients was incubated with cytidine for 16 h (overnight) at 37°C. After protein precipitation, supernatant containing produced ammonia was collected and then treated with both phenol and sodium hydrochloride. The concentration of produced indophenol was assayed by spectrophotometry at 620 nm. Results of CDA activity were expressed in U per mg of protein with 1 U = 4.10–3 μmol of ammonium released per min and per ml of serum. The lower limit of quantification of CDA activity assay was 0.5 U. Six replicates of CDA assessment in serum from one healthy volunteer were assayed the same day for the intraday experiment, and three replicates were assayed on 6 different days for the interday experiment. The intra‐ and interday precision of the analytical assay were 3.2% and 6.2%, respectively.

CDA genotyping analysis

A 7‐ml blood sample was collected in ethylenediaminetetraacetic acid tube at the inclusion time. Genomic DNA was extracted from peripheral blood leucocytes using the Maxwell 16 LEV Blood DNA Kit (Promega Corporation, Charbonnières‐les‐Bains, France) according to the recommendations of the manufacturer in the biochemistry laboratory (European Georges Pompidou hospital, Paris). The most described single nucleotide polymorphisms (SNPs) in CDA gene from Caucasian population have been genotyped: two non‐synonymous c.79 A > C (rs2072671) and c.435 C > T (rs1048977) substitutions and three SNPs located in the CDA promoter region, c.‐33_‐31delC (rs3215400), c.‐92A > G (rs602950) and c.‐451C > T (rs532545). The nonsynonymous c.208 G > A (rs60369023), known to both decrease CDA activity and gemcitabine clearance, is not detected in Caucasian population 9, 30. Therefore, the analysis of this variant was not performed in our cohort.

CDA exons, splice junctions and a portion of the 5′‐flanking regions, were amplified using polymerase chain reaction. Primers were designed to hybridize within introns, in the 5′‐flanking regions or in the 3′‐untranslated regions of the terminal exon. The cycling conditions and primers are listed in Supplementary data (Supplemental Table S2). The sequencing was performed on an ABI Prism Genetic Analyzer System 9700 (Applied Biosystems, Life Technologies SAS, Courtabœuf, France).

CDA activity and gemcitabine toxicity

The objective was to assess the relationship between baseline serum CDA activity and the occurrence of severe gemcitabine‐related adverse events (grade ≥3) in 56 patients regardless of tumour location. Patients could receive gemcitabine in monotherapy or in combination with other chemotherapies. At each cycle, adverse events were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI‐CTCAE) version 3.0. In case of grade ≥3 toxicity, the treatment was discontinued until the adverse events returned to grade <2 or to baseline. Subsequent dose reductions were left to the discretion of the attending physicians. Gemcitabine administration was continued until disease progression or intolerable toxicity.

Longitudinal monitoring of CDA activity in pancreatic cancer patients

CDA activity was assayed in six patients with metastatic pancreatic cancer, treated with gemcitabine as a single agent following a weekly schedule [1000 mg m^–2 Day (D)1, D8 and D15 every 4 weeks]. Serum CDA activity was assessed at baseline and just before the first infusion of cycles 2 (D28), 3 (D56) and 4 (D84).

Statistical analysis

CDA activity, neutrophil and CRP values were log‐transformed before all analysis, to handle the high skewness they presented. Unless otherwise specified, all analyses were conducted through the linear model framework (one‐way analysis of variance, linear regression, multiple regressions). Conditions to apply this framework, namely normality and homoscedasticity, were assessed graphically before considering the results of the analysis. Univariate analyses were mainly conducted as a variable selection step for the multivariate model: those variables with P < 0.1 were included in the final model. The multivariate model was analysed using likelihood ratio tests for nested models, using P < 0.05 to consider a variable significant. Multicolinearity issues were checked on the selected predictors. Since some of them were highly correlated, a complementary analysis using partial least‐squares regression (PLS) was conducted to have robust results. PLS regression was cross validated using a leave‐one‐out algorithm, yielding a model with three components. P‐values of regression coefficients were computed using approximate t‐tests based on jack‐knife variance estimates. A χ² test was used to compare the incidence of grade ≥3 adverse events in gemcitabine‐treated patients according to their deficiency of serum CDA activity (<2 U mg^–1). Longitudinal follow‐up data were analysed using a linear mixed effects model, with patient as the random effect (package lmer4 for R) and log (CRP) as a covariate. Tests were done using asymptotic likelihood ratio tests for nested models. All analyses were run using R version 3.3.1 and additional packages lme4 and pls.

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 31, and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 32.

Results

Population's characteristics

Three hundred and four patients were included in the study; 121 patients were excluded for different reasons: written consent not obtained (n = 103), no CDA phenotyping (n = 17), ongoing chemotherapy (n = 1). Clinical, biological and demographic characteristics of the analysis cohort (n = 183) are summarized in Table 1. Planned chemotherapies are described in supplementary table S3.

Table 1.

Clinical and demographic characteristics of the analysis cohort

	Overall population (n = 183)	Gemcitabine‐treated population (n = 56)
Median age (years)	65 [55–75]	69 [57–76]
Gender
female (%)	72 (39)	19 (34)
male (%)	111 (61)	37 (66)
Active smoking, n (%)	31 (17)	10 (18)
ECOG PS, n (%)
0–1	131 (72)	33 (59)
≥2	52 (28)	23 (41)
Body mass index (kg/m 2 )	23.7 [21.3–28.0]	22.6 (20.1–26.9)
Malnutrition a , n (%)
none	113 (62)	27 (48)
mild	47 (26)	19 (34)
severe	23 (12)	10 (18)
Inflammatory syndrome b , n (%)	59 (32)	22 (39)
Liver disease, n (%)	5 (3)	7 (12)
Metastatic disease, n (%)	124 (68)	41 (73)
Liver metastasis, n (%)	29 (16)	13 (23)
Primary tumour location, n (%)
lung	31 (17)	5 (9)
prostate	23 (13)	0 (0)
soft tissues	22 (12)	6 (11)
urinary tract	17 (9)	16 (29)
colon–rectum	16 (9)	0 (0)
pancreas	10 (5)	9 (16)
other	64 (35)	20 (36)
Biological characteristics
Leucocytes (cells μl –1 )	7270 [5615‐9575]	7845 [5345–10 090]
Lymphocytes (cells μl –1 )	1520 [1090‐1995]	1515 [0.995–1955]
Neutrophils (cells μl –1 )	4704 [3440‐6485]	4930 [3610–8015]
AST (U l –1 )	26 [20–33]c	27 [20–37]
ALT (U l –1 )	24 [17–35]c	24 [18–45]
GGT (U l –1 )	40 [24–84]c	49 [32–106]
ALP (U l –1 )	90 [6–138]c	94 [76–144]
Total bilirubin (μmol l –1 )	6.8 [5.3–9.1]	7,2 [5,6‐10,4]
Albumin (g l –1 )	42 [39–44]d	42 [40–44]
C‐reactive protein (mg l –1 )	5.6 [1.8–23.8]	6,55 [1.9–34,5]
Creatinine (μmol l –1 )	73 [61–86]	73 [61–85]

ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; ECOG PS, Eastern Cooperative Oncology Group performance status; GGT, gamma‐glutamyltransferase

^aMalnutrition was defined by BMI < 18.5 kg m^–2 and/or albumin <35 g l^–1

^bInflammatory syndrome was defined as increased levels of at least one inflammatory biomarker: CRP, absolute neutrophil count or ferritin

^cOne missing value

^dThree missing values

Quantitative variables are expressed as median [interquartile range]

CDA phenotype distribution

Figure 1 presents the distribution of CDA activity in the analysis cohort (n = 183). Median CDA activity was 3.97 U mg^–1 (range 1.53–15.49 U mg^–1, 25^th–75^th percentile: 3.22–5.03 U mg^–1). Nine patients (5%) exhibited a <2 U mg^–1 CDA activity, but none with CDA activity <1.3 U mg^–1. Finally, 30 patients (16%) were characterized as ultrarapid metabolizers (UM) (CDA activity > 6 U mg⁻¹).

Figure 1.

Distribution of cytidine deaminase (CDA) enzymatic activity. Black line: CDA enzymatic activity distribution (log); blue dotted line: log‐normal distribution; green: arithmetic mean ± standard deviation; box plot: median and quartiles; grey dots and dashes: individual CDA enzymatic values

CDA genotype distribution

Variant allele frequencies in the cohort study (> 90% of Caucasian patients) were in accordance with those previously reported in the literature for Caucasian population (Table 2) 33. Genotype distributions were consistent with the Hardy–Weinberg equilibrium. The most frequent polymorphism observed in the analysis cohort (n = 178) was c.‐33_‐31delC in 82% of patients (n = 147) including 88 WT/MT (wild‐type allele/mutated allele) patients (49%) and 59 MT/MT patients (33%).

Table 2.

CDA genotyping

Single nucleotide polymorphism	n	rs	Genotype			Minor allele			HWE P‐value
			Wild‐type n (%)	Heterozygous mutation n (%)	Homozygous mutation n (%)	Type	Number	Frequency
79A > C	178	rs2072671	79 (44)	75 (42)	24 (14)	MT	123	35	0.6382
435C > T	178	rs1048977	87 (49)	69 (39)	22 (12)	MT	113	32	0.8405
‐451G > A	181	rs532545	85 (47)	76 (42)	20 (11)	MT	116	32	0.3707
‐92A > G	177	rs602950	82 (46)	76 (43)	19 (11)	MT	114	32	0.1755
‐31Del/InsC	178	rs3215400	31 (18)	88 (49)	59 (33)	WT	150	42	0.1466

MT: mutated allele; WT: wild‐type allele; HWE: Hardy–Weinberg equilibrium

Factors for the interindividual variability in CDA activity

Results of the univariate analysis are shown in Table 3. CDA activity appeared significantly positively associated with Eastern Cooperative Oncology Group performance status, mild or severe malnutrition, inflammatory syndrome, leucocyte count, neutrophil count, low albumin level, CRP and c.‐33_‐31delC. In multivariate analysis, only high neutrophil count (P < 0.0001) and severe malnutrition (P = 0.0278) remained positively associated with CDA activity (Table 3). For instance, for a given neutrophil count, a 19.5% increase in CDA activity is expected [95% confidence interval (95%CI) 2–40%] in patients with severe malnutrition. For a given nutritional status, a 40% increase in the neutrophil count is expected to give a 10% increase in serum CDA activity (95%CI 6.3–14.8%). Finally, patients with the c.‐33_‐31delC SNP had slightly higher CDA activities (+13.7%; 95%CI −0.1 to 29.5%; P = 0.052). Overall, when considered separately, neutrophil count, malnutrition and c.‐33_‐31delC accounted for 24.8, 8.9 and 4.1% interindividual variability in CDA activity, respectively, whereas the complete multivariate model accounted for 33% of the CDA activity variability. The PLS model with three components retained a significant association with the CDA activity for neutrophil count (P < 0.0001), severe malnutrition (P = 0.0224) and c.‐33_‐31delC (P = 0.0248) and accounted for 32% of the CDA activity. Finally, multivariate analysis (PLS analysis or classical multiple regression) showed that CRP was not statistically associated with CDA activity.

Table 3.

Univariate and multivariate analysis (linear model)

Covariable			Univariate analysis		Explained part of interindividual variability	Multivariate analysis
		n	95%CI	P‐value		95%CI	P‐value
Sex	Female	72	reference
	Male	111	−5.09% [−15.30; 6.34]	0.3657	0.45%
Age		183	−0.15% [−0.50; 0.21]	0.4156	0.37%
BMI		183	−0.19% [−1.30; 0.92]	0.7315	0.07%
Active smoking	No	152	Reference
	Yes	31	−1.27% [−14.89; 14.53]	0.8651	0.01%
ECOG PS	0–1	131	Reference			Reference
	≥ 2	52	+9.91% [0.83; 19.81]	0.0319a	2.52%	−2.48% [−10.14; 5.83]	0.54
Malnutrition	No	113	Reference			Reference
	Mild	47	+14.26% [0.82; 29.50]	0.0370a	8.87%	−0.07% [−11.66; 13.03]	0.99
	Severe	23	+39.69% [18.44; 64.75]	<0.0001a		+19.45% [1.98; 39.91]	0.028a
Inflammatory syndrome	No	124	Reference			Reference
	Yes	59	+32.55% [18.54; 48.23]	<0.0001a	12.03%	+10.16% [−9.14; 33.55]	0.32
Liver disease	No	178	Reference
	Yes	5	−22.04% [−44.49; 9.502]	0.1499	1.14%
Metastatic disease	No	59	Reference
	Yes	124	+5.32% [−6.49; 18.62]	0.3913	0.41%
Liver metastasis	No	154	Reference
	Yes	29	+7.03% [−8.08; 24.63]	0.3795	0.43%
Primary tumour location	Lung	31	Reference
	Colon–rectum	16	−9.59% [−28.31; 14.03]	0.3926	3.36%
	Prostate	23	−18.54% [−33.80; 0.25]	0.0528
	Soft tissues	22	−7.10% [−24.71; 14.63]	0.4900
	Urinary tract	17	−10.52% [−28.73; 12.35]	0.3365
	Pancreas	10	−8.65% [−30.55; 20.17]	0.5159
	Others	64	−3.189% [−17.91; 14.17]	0.6987
Leucocytes (per 2000 μl –1 increase)		183	10.17% [7.13; 13.30]	<0.0001a	20.47%
Lymphocytes (per 2000 μl –1 increase)		183	−3.38% [−11.63; 5.65]	0.45	0.32%
Neutrophils (per 2000 μl –1 increase)		183	13.47% [9.86; 17.19]	<0.0001a	24.78%	+34.33% [19.83; 50.58]	<0.0001a
AST		182	−0.083% [−0.39; 0.23]	0.60	0.15%
ALT		182	−0.079% [−0.28; 0.12]	0.44	0.34%
ALP		182	0.017% [−0.02; 0.05]	0.31	0.57%
GGT		182	0.014% [−0.01; 0.04]	0.33	0.52%
Total bilirubin		183	−0.10% [−0.43; 0.23]	0.56	0.19%
Albumin		180	−2.61% [−3.65; −1.56]	<0.0001a	11.66%	−1.05% [−2.28; 0.20]	0.099
CRP		183	0.31% [0.16; 0.46]	<0.0001a	8.24%	−1.57% [−7.57; 4.82]	0.62
Creatinine		183	−0.12% [−0.35; 0.10]	0.28	0.63%
−451G > A	WT	85	Reference
	MT	96	+7.87% [−3.51; 20.60]	0.18	0.99%
−92A > G	WT	82	Reference
	MT	95	+5.43% [−5.89; 18.10]	0.36	0.48%
–31Del/InsC	WT	31	Reference			Reference
	MT	147	+22.53% [5.84; 41.86]	0.0068a	4.09%	+13.74% [−0.097; 29.49]	0.052
79A > C	WT	79	Reference
	MT	99	−2.97% [−13.43; 8.75]	0.60	0.15%
435C > T	WT	87	Reference
	MT	91	−1.06% [−11.67; 10.83]	0.85	0.019%

95%CI: 95% confidence interval; ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; CRP, C‐reactive protein; ECOG PS, Eastern Cooperative Oncology Group performance status; GGT, γ‐glutamyl transpeptidase; MT: mutated allele; WT: wild‐type

^a: statistically significant

Relationship between CDA activity and severe toxicity of gemcitabine

Among the 56 patients treated with gemcitabine (Table 1), 11 patients (19.6%) developed severe toxicities which were mainly haematological: five patients (8.9%) experienced grade 3 or 4, neutropenia and four patients (7.1%) a grade 3 thrombocytopenia. One patient developed a hepatotoxicity associated with an acute renal failure; another patient a tubular nephropathy potentially attributable to gemcitabine. Given that no patient exhibited CDA activity <1.3 U mg^–1, we considered that patients with a CDA activity <2 U mg^–1 were deficient. Among 56 patients, eight (14%) presented a deficiency of serum CDA activity. In this group of patients, three (37.5%) experienced severe toxicities and eight (16.7%) in the group of patients with CDA activity >2 U mg^–1. There was a trend to a worse tolerance of gemcitabine in patients with low CDA activity (<2 U mg^–1); however, this result did not reach statistical significance (P = 0.16).

Intraindividual variability in CDA activity in pancreatic cancer patients

Six metastatic pancreatic cancer patients (four women, median age 64 years, range 46–76 years) were treated with 1000 mg m^–2 gemcitabine monotherapy at a 1535 mg median weekly dose (range: 1280–2000). All patients were treated for at least five cycles. Figure 2 displays the CDA activities and neutrophil counts of each of these patients during the first 90 days of treatment course.

Figure 2.

Intraindividual variability in cytidine deaminase (CDA) activity and neutrophil count in six pancreatic cancer patients treated with gemcitabine. Blue and pink lines represent neutrophil count and CDA activity, respectively

For all six patients, CDA activity decreased from baseline to the end of the first gemcitabine cycle (median change, −1.6 U mg^–1; range, −2.9 to −0.9; P = 0.03, Wilcoxon rank signed test for paired samples), whereas change in neutrophil count and CRP did not show any significant variation (P = 0.1563 and P = 0.2807, respectively). There was no significant relationship between CDA and neutrophil count variations (P = 0.2, Spearman's ranks correlation coefficient test).

Interestingly, between the end of the first cycle and the end of the last cycle, CDA activity significantly decreased (P = 0.03), while significant variation in neither CRP level (P = 0.75) nor neutrophil count (P = 0.99) were observed in these patients. During the three cycles of gemcitabine therapy, mean intraindividual variability in CDA activity and neutrophil count was 16 ± 11% and 19 ± 9%, respectively.

When considering simultaneously the baseline and gemcitabine cycles CDA activities, a relationship between CDA activity and neutrophil count (P = 0.01) was observed, with an expected increase of 4.0% in CDA activity (95%CI 1.0–6.4%) per 40% increase of neutrophil count. In contrast, no significant association was found between CDA activity and CRP (P = 0.6843).

Discussion

The measurement of baseline serum CDA activity in cancer patients has been proposed as a promising circulating biomarker for nucleoside analogues toxicity and efficacy 8, 34. However, contradictory results have been reported. This might be explained by patient selection bias in the context of the high interindividual variability in CDA serum activity. Thus, the identification of factors contributing to this variability becomes a mandatory step before the use of serum CDA activity in daily clinical practice. Here we show that neutrophil count and nutritional status are two factors of this interindividual variability in 183 unselected cancer patients.

In the present cohort, the mean serum CDA activity was 4.32 ± 1.80 U mg^–1, which is consistent with the results previously reported in the literature. Additionally, our study confirms the large interindividual variability in serum CDA activity 10, 12, 13. The threshold value of CDA activity to define poor metabolizers (PM) depends on chemotherapy treatment: <1.3 U mg^–1 for gemcitabine and <2 U mg^–1 for cytarabine 10, 35. In the present study, no patients with CDA activity <1.3 U mg^–1 were identified while only 5% of patients exhibited CDA activity below 2.0 U mg^–1. Intriguingly, 48% of patients with CDA activity <2 U mg^–1 have been reported in a cohort of patients with acute myeloid leukaemia (AML) 35. The pathophysiology of AML, characterized by a differentiation block in the myeloid lineage leading to the accumulation of immature blood cells in bloodstream. The CDA activity is known to be significantly lower (3.55–14.2‐fold) in immature blood cells compared to mature granulocytes 36, which could contribute to a higher rate of PM in patients with AML. In the present study, we observed a trend towards higher incidence of severe toxicities in gemcitabine‐treated patients with CDA activity below 2 U mg^–1 (i.e. 37.5 vs. 16.7%), although the small sample size prevents statistical analysis to be fully conclusive here. Nevertheless, the data are in line with the hypothesis that CDA status could be a predictive marker for severe toxicities in gemcitabine‐treated patients. By contrast, 16% of patients were considered as UM (i.e. CDA activity ≥ 6.0 U mg^–1), which is consistent with a previously published analysis reporting 14% of UM in 40 pancreatic cancer patients 15. Surprisingly, a recent study conducted by the same team reported 4.6% of PM and only 0.9% of UM in 120 selected pancreatic cancer patients (FFCD‐1004 clinical trial; NCT01416662) 34. They did not show any relationship between baseline serum CDA activity and clinical outcomes (toxicity, efficacy), probably because of a small interindividual variability related to the selection of patients (i.e. patients fit for adjuvant therapy after curative pancreatic surgery). This observation supports the recommendation that pharmacodynamic studies focusing on serum CDA activity should be further conducted among patients from the real world rather than selected patients. Finally, we did not investigate the association between CDA activity and clinical outcomes of anticancer agents other gemcitabine for two reasons: (i) because of the multiplicity of chemotherapy regimens with small sample size (see supplementary table S3); and (ii) to avoid addressing multiple statistical tests among samples that might impair the α risk.

Serum CDA activity was proposed in the 1980s as a biomarker of activity for several inflammatory diseases such as rheumatoid or psoriatic arthritis 26, 27. As far as we know, the present study is the first to show a significant association between serum CDA activity and neutrophil count in cancer patients. Neutrophil count could explain almost 25% of the interindividual variability in serum CDA activity, when considered as the only predictor. Patients exhibiting a high neutrophil count are likely to have a greater serum CDA activity and 40% of UM exhibited a neutrophil count beyond the upper limit range (>7000 cells μl^–1). Neutrophils largely express CDA and are short‐lived with a circulating half‐life of 6–8 h 37. One hypothesis would be that CDA could be released from neutrophils during their breakdown. Interestingly, it has been suggested that CDA activity follows a circadian rhythm 38. Increased CDA activity was observed in patients with physical activity compared to patients at rest. In this context, one cannot exclude that the positive relationship between CDA activity and neutrophils was at least partly due to exercise resulting in mobilization of the marginated neutrophils. Our results suggest that this CDA release would be strongly enhanced during inflammatory syndrome (P < 0.001 in univariate analysis). Half‐life of gemcitabine is approximately 30 min, which suggests that circulating CDA could be able to form the inactive compound 2′,2′‐difluorodeoxyuridine in the bloodstream as well as in liver and kidneys 39. In this context, CDA released by the neutrophils could contribute to the metabolism of gemcitabine and other nucleoside analogues. According to this hypothesis, our observation would be clinically meaningful since patients who receive gemcitabine, cytarabine or azacytidine while having chemotherapy‐induced neutropenia could experience increased toxicities.

Interestingly, once adjusted on neutrophil count, high serum CDA activity was also significantly associated with severe malnutrition. The clinical interest of this finding is supported by the observation that patients with severe malnutrition are more likely to develop severe capecitabine‐related toxicities 40. Indeed, high activity of CDA combined with malnutrition could be associated with plasma 5‐fluorouridine overexposure related to an increased CDA‐through clearance converting capecitabine to 5‐fluorouridine 41. These results suggest that UM patients should be identified before prescription of capecitabine.

The c.‐33_‐31delC SNP was the only one associated with serum CDA activity, although the statistical significance was borderline in the multivariate analysis (P = 0.052). In univariate analysis, an increase of 19% and 25% in serum CDA activity was observed in c.‐33_‐31delC homozygous and heterozygous patients compared to wild type patients, respectively. As far we know, the association between the c.‐33_‐31delC SNP and CDA phenotype has never been reported yet. However, Caronia et al. documented that the deletion c.‐33_‐31delC in the promoter of CDA enhances the CDA expression 23. Haplotype CDA*2B containing the c.‐33_‐31delC SNP was associated with a significant increase in CDA activity 33. Furthermore, c.‐33_‐31delC SNP has been associated with severe capecitabine‐related toxicities and cytarabine efficacy 23, 42. Taken together, these functional, genetic and clinical data results support the increased activity of CDA in c.‐33_‐31delC mutated patients observed in our cohort. One limit of our CDA genotyping analysis is the small number of patients given the known allele frequencies. In this context, further clinical investigations are required to confirm the clinical relevance of increased CDA activity in c.‐33_‐31delC mutated patients.

As far as we know, the present study is the first to explore the intraindividual variability in serum CDA activity in cancer patients treated with a CDA‐metabolized drug. The characteristics of patient recruitment in our centre led us to include pancreatic cancer patients treated with gemcitabine as a single agent. It should be noted that the intraindividual variability in serum CDA activity has been explored in rheumatoid arthritis patients before and after withdrawal of nonsteroidal anti‐inflammatory treatment 43. Interestingly, we observed a significant decrease in CDA activity during the longitudinal follow‐up of six pancreatic cancer patients treated with gemcitabine. The impact of neutrophil count on baseline serum CDA activity previously observed in our 183 cancer patients was confirmed in these patients over the treatment course. This result suggests that baseline serum CDA activity alone could not correctly predict the behaviour of CDA activity over the treatment course. Moreover, this suggests that the total clearance of gemcitabine could be decreased over the time in pancreatic cancer patients. This could have a significant clinical impact, especially in patients with low baseline CDA activity (i.e. CDA activity <2.0 U mg^–1) who could experience an overexposure to gemcitabine and therefore severe drug‐related toxicities. Our results suggest that decrease in CDA activity over time could be at least partly related to relative neutropenia induced by gemcitabine itself. In this hypothesis, the more gemcitabine or other associated drug is haematotoxic, the more gemcitabine exposure would be increased. One might also hypothesize that low CDA activity is responsible for an increased haematotoxicity of gemcitabine and consequently lower its own activity. In addition, we cannot exclude a direct effect of gemcitabine on CDA activity by mean of enzymatic inhibition. However, our exploratory approach presents two limitations to firmly conclude about this issue: the small size of the cohort (n = 6) and the lack of pharmacokinetic data for gemcitabine and its metabolites. Nevertheless, this exploratory analysis paves the way to investigate the relationship between clinical outcomes and serum CDA activity over the time in a context of a tailored‐approach of nucleoside analogue dosing.

The present study highlights that elevated neutrophil count and malnutrition are significantly correlated with increased serum CDA activity in cancer patients. Thus, our work suggests that serum CDA activity should be interpreted along with the inflammation and malnutrition status not to underestimate the frequency of CDA‐deficient patients. These findings suggest that pharmacologists and physicians should take account for these parameters in their interpretation of pretherapeutic CDA activity in daily clinical practice. Finally, further research should focus on the intraindividual variability in CDA activity over time, an important issue in the development of personalized strategies for chemotherapy dosing.

Competing Interests

There are no competing interests to declare.

The authors wish to thank Sylvie Lacassagne for her assistance in writing this manuscript.

Contributors

R.C., E.C., C.J., D.D., I.A., J.A. and B.B. wrote the manuscript; R.C., A.C., F.G., J.A. and B.B. designed the research; R.C., L.H.P., V.J., O.H., A.J., C.N., A.TS., A.B., A.C., M.V., M.TM., J.Q., F.G., J.A. and B.B. performed the research; E.C., C.N., D.D., I.A. and B.B. analysed the data; LH.P., V.J., C.N., A.TS., M.V. and B.B. contributed new reagents/analytical tools.

Supporting information

Table S1 Definition of mild and severe malnutrition

Table S2 Primer sequences used to perform amplifications

Table S3 Planned chemotherapies in 183 patients

Cohen R., Preta L. H., Joste V., Curis E., Huillard O., Jouinot A., Narjoz C., Thomas‐Schoemann A., Bellesoeur A., Tiako Meyo M., Quilichini J., Desaulle D., Nicolis I., Cessot A., Vidal M., Goldwasser F., Alexandre J., and Blanchet B. (2019) Determinants of the interindividual variability in serum cytidine deaminase activity of patients with solid tumours, Br J Clin Pharmacol, 85, 1227–1238. doi: 10.1111/bcp.13849.