Biochemical Journal

Research article

Poly(ADP-ribose) polymerase-1 (PARP-1) pharmacogenetics, activity and expression analysis in cancer patients and healthy volunteers

Tomasz Zaremba, Huw D. Thomas, Michael Cole, Sally A. Coulthard, Elizabeth R. Plummer, Nicola J. Curtin

Abstract

There is a wide inter-individual variation in PARP-1 {PAR [poly(ADP-ribose)] polymerase 1} activity, which may have implications for health. We investigated if the variation: (i) is due to polymorphisms in the PARP-1 gene or PARP-1 protein expression; and (ii) affects patients' response to anticancer treatment. We studied 56 HV (healthy volunteers) and 118 CP (cancer patients) with supporting in vivo experiments. PARP activity ranged between 10 and 2600 pmol of PAR/106 cells and expression between 0.02–1.55 ng of PARP-1/μg of protein. PARP-1 expression correlated with activity in HV (R2=0.19, P=0.003) and CP (R2=0.06, P=0.01). A short CA repeat in the promoter was significantly associated with increased cancer risk [OR (odds ratio), 5.22; 95% CI (confidence interval), 1.79–15.24]. PARP activity was higher in men than women (P=0.04) in the HV. Male mice also had higher PARP activity than females or castrated males. Oestrogen supplementation activated PARP in PBMCs (peripheral blood mononuclear cells) from female mice (P=0.003), but inhibited PARP-1 in their livers by 80%. PARP activity and expression were not dependent on the investigated polymorphisms, but there was a modest correlation of PARP activity with expression. Studies in the HV revealed sex differences in PARP activity, which was confirmed in mice and shown to be associated with sex hormones. Toxic response to treatment was not associated with PARP activity and/or expression.

  • chemotherapy
  • DNA repair
  • poly(ADP-ribose) polymerase (PARP)
  • poly(ADP-ribose) polymerase activity
  • poly(ADP-ribose) polymerase expression
  • radiotherapy

INTRODUCTION

PARP-1 {PAR [poly(ADP-ribose)] polymerase-1} is involved in DNA repair, genomic stability, transcription control, cell death and proliferation (reviewed in [1,2]). Binding of PARP-1 at DNA breaks activates the enzyme to cleave NAD+ and create long homopolymers of ADP-ribose attached to both PARP-1 itself and histone tails at the vicinity of the break, thereby ‘flagging’ the damage to the repair machinery. PARP-1-knockout mice, and the cells derived from them, are hypersensitive to DNA methylating agents, topoisomerase I poisons and ionizing radiation. These agents are used in the treatment of cancer, and PARP-1 inhibitors increase their anticancer activity (reviewed in [1,2]). Paradoxically, PARP activity can also promote cell death in non-replicating normal cells that are not so dependent on rapid DNA repair. In such cells and tissues, a burst in reactive oxygen species formation following ischaemia–reperfusion injury, infection and inflammation leads to DNA breaks that activate PARP-1, resulting in rapid and catastrophic NAD+ and ATP depletion and subsequent cell death (reviewed in [3]). Clearly, PARP activity has implications in human health and disease and response to anticancer therapy. Large inter-individual differences in PARP activity in PBMCs (peripheral blood mononuclear cells) have been reported in both HV (healthy volunteers) and CP (cancer patients) [47]. High PARP activity may promote DNA repair and genomic stability in normal cells as well as cancer cells, thus can lead to resistance to DNA-damaging anticancer treatment. However, low PARP activity may lead to reduced pro-inflammatory mediators, tissue damage, necrosis and reperfusion injury.

Little is known about the potential underlying mechanisms responsible for the variation in PARP-1 activity and expression. There are at least 60 reported SNPs (single nucleotide polymorphisms) in the PARP-1 sequence (http://snp500cancer.nci.nih.gov). One of these polymorphisms in the promoter region is a microsatellite polymorphic DNA fragment, consisting of a variable number of CA repeats [8] that may facilitate transcription from the promoter via the formation of DNA quadruplex structures [9]. Furthermore, the CA microsatellite is located close to the binding site of the transcription factor Yin Yang 1, and this may also contribute to the regulation of transcription [10,11]. The common 2444T>C SNP (at a frequency of 5–33%), resulting in an amino acid substitution, V762A, in the PARP-1 catalytic domain, has been reported to reduce PARP-1 catalytic activity by 30–40% and to be associated with various cancers [4,1215].

Numerous studies suggest a correlation between PARP activity and age. A positive correlation between specific PARP activity and mean maximal lifespan in 13 mammalian species as well as a decrease in PARP-1 activity with age in humans and rats was reported previously [16]. In contrast, enhancement of PARP activity was reported in brains of old adult animals compared with young controls [17] and in lymphoblastoid cell lines derived from centenarians [7].

Patients vary in their toxic and therapeutic response to treatment owing to several different factors, and pharmacogenetics may be used to predict toxicity and response, allowing more tailored drug treatment. Previous clinical trials with PARP inhibitors indicate that suppression of PARP activity can have a profound effect on chemotherapy-induced toxicity [18] as well as the efficacy of chemotherapy [19]. An understanding of the genetic determinants of PARP activity and its relation to patients' response was investigated in the present study.

Our aim was to further evaluate the inter-individual differences in PARP activity and determine the underlying mechanisms responsible for the variation in terms of PARP-1 protein expression, polymorphisms in the PARP-1 gene and demographic factors such as age and sex. We investigated the underlying mechanisms by measuring PARP-1 polymorphisms, expression and activity in PBMCs from 118 CP and 56 HV, with supporting in vivo studies. We also studied if PARP activity contributes to patients' response to treatment in terms of toxicity and if particular types of malignancy are associated with higher or lower PARP activity.

MATERIALS AND METHODS

Chemicals

β-Oestradiol 17-valerate and all routine chemicals and tissue culture reagents were supplied by Sigma–Aldrich unless otherwise stated. AG014699 was a gift from Dr Zdenek Hostomsky (Pfizer Oncology, La Jolla, CA, U.S.A.).

Cell culture

Chronic myelogenous leukaemia K-562 cells obtained from A.T.C.C. (CCL-243) were cultured in RPMI 1640 medium with 10% fetal bovine serum and 1% antibiotic/antimycotic at 37 °C in an atmosphere of 5% CO2 in air. Cells were confirmed Mycoplasma-negative by regular testing (Mycoalert; Cambrex).

Hormonal manipulation in mice

The animal study was conducted in accordance with national law and institutional guidelines under a protocol approved by the local ethics committee. CD-1 mice, 8–10 weeks of age (Charles River Laboratories), were treated as follows: male untreated controls (n=9), castrated untreated males (n=9), castrated males treated with 4 mg of β-oestradiol 17-valerate per mouse dissolved in corn oil (n=15) by a single intramuscular injection on the day of castration, untreated females (n=9) and females (n=15) treated with β-oestradiol 17-valerate as above. Animals were killed 6 days later and blood from controls (n=3) and treated animals (n=5) was pooled prior to collection of PBMCs. Livers from control female mice (n=3) and from oestradiol-treated female mice (n=3) were also collected for analysis and stored at −80 °C.

Human subjects

The research protocol for the PARP clinical study was approved by the local ethics committee, and was carried out in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki as revised in 2000. This protocol and the associated patient information sheet and consent form comply with the guidance contained in the Medical Research Council Operational and Ethical Guidelines: Human Tissue and Biological Samples for use in Research (April 2001) and The European Directive on the conduct of Medical Research (April 2004). The present study was conducted according to ICH (International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use) good clinical practice guidelines and laboratory work according to ICH good laboratory practice. The present study included CP newly diagnosed with solid tumours who were referred to the NCCT (Northern Centre for Cancer Treatment) between February 2007 and December 2008 and HV. Subjects supplied a blood sample (10–20 ml) and data including date of birth, sex, ethnicity, weight, type of diagnosed disease, stage and grade, treatment, co-medication, co-morbidities and response to treatment (CP) or data on their sex, age, weight and ethnicity (HV). The demographic characteristics of the HV and CP in this study are listed in Table 1.

View this table:
Table 1 Baseline characteristics of the eligible group of participants

For age and weight, mean values±S.D. are given, with range in brackets and median value in curly brackets (age only). Information on age was available for n=156 and for weight n=139.

Assessment of toxicity in patients undergoing anticancer treatment

Toxicity after the first cycle of chemotherapy or the first course of radiotherapy or concurrent radiotherapy and chemotherapy was graded according to The National Cancer Institute Common Terminology Criteria for Adverse Events version 3.0 (CTCAE). Assessment of neutropenia or other myelotoxicity was based upon blood analysis before second cycle of treatment. The analysis of toxicity was based on a comparison of the rates of grade 3 and greater toxicity.

Genotyping

DNA was isolated directly from blood using the Blood Mini Kit (Qiagen) according to the manufacturer's instructions and genotyped for the 2444T>C SNP by pyrosequencing using the PSQ96 system (Pyrosequencing) and for CA microsatellite capillary using the electrophoresis system CEQ8000 (Beckman Coulter) as previously described [20].

Western blot analysis

Briefly, the cell pellet was lysed in 100 μl of Laemmli Sample buffer with 1× Halt protease inhibitor cocktail (Thermo Fisher Scientific), sonicated on ice for 10 s and heated in loading dye containing 2-mercaptoethanol and Bromophenol Blue at 95 °C for 5 min. Lysates (30 μg of protein per lane) were run on Tris/HCl 5–20% polyacrylamide gels (Bio-Rad) along with purified recombinant PARP-1 immunoblotting standard (0–40 ng: Enzo Life Sciences) at 100 V for 2 h and transferred for 1 h at 4 °C to a nitrocellulose membrane (Hybond-C; Amersham) on a Criterion electrophoresis and blotting apparatus (Bio-Rad). After blocking for 1 h in PBS-MT (PBS plus 5% non-fat dried skimmed milk powder and 0.5% Tween 20), the membrane was incubated overnight at 4 °C with an anti-PARP-1 C2–10 primary antibody (1:2000 in PBS-MT; Trevigen) washed three times in PBS-T (PBS+ 0.5% Tween 20), and then incubated with the HRP (horseradish peroxidase)-linked secondary goat anti-mouse antibody (1:1000 in PBS-MT; Dako), washed again for 1 h in PBS-T (PBS and 0.5% Tween 20) and dried. The protein was visualized with the ECL Plus detection kit (GE Healthcare) using the manufacturer's protocol followed by chemiluminescence detection using a Fuji LAS3000 with imaging software (Fuji LAS Image version 1.1; Raytek). PARP-1 expression was quantified by reference to the recombinant PARP-1 standard curve. This assay was validated to GCLP (Good Clinical Laboratory Practice) standards for evaluation of patient samples (E. Mulligan and T. Zaremba, unpublished work). Validation studies showed that loading the lysate in duplicate with protein determination gave more reliable data than use of a loading control such as GAPDH (glyceraldehyde-3-phosphate dehydrogenase) or β-actin. Additionally, using a purified PARP-1 standard and a quality control sample (protein extract from K-562 cells) assured the quality of transfer and allowed the most precise protein quantification.

PARP activity assay

Total stimulatable PARP activity was measured by modification of a previously described method [6] validated to the GCLP standards and used as a pharmacodynamic endpoint for clinical trials [18]. This assay measures PARP activity that has been maximally stimulated by a double-stranded oligonucleotide in the presence of excess NAD+, thereby eliminating error due to variable activation of the enzyme by DNA damage accidentally introduced during processing. Quality control samples of L1210 cells were included in each assay. As part of the validation of this assay, the day-to-day variation between samples from the same individual was measured in independent experiments. In eight individual HV, the mean maximum variation in PARP activity, measured on three different days, was 1.5±0.2-fold. PARP activity was measured in triplicate samples of 104 digitonin-permeabilized cells in a reaction mixture containing 350 μmol/l NAD+ and 10 μg/ml oligonucleotide (CGGAATTCCG) (Europrim) in a reaction buffer of 100 mmol/l Tris/HCl and 120 mmol/l MgCl2 (pH 7.8) in a final volume of 100 μl for 6 min at 26 °C. After blotting on to a nitrocellulose membrane (Hybond-N; Amersham), the PAR was detected following incubation with the primary anti-PAR 10H antibody (1:1000) then with HRP-conjugated goat anti-mouse secondary antibody (1:1000; Dako) and finally ECL reaction and chemiluminesence detection as described above. Results were expressed relative to the number of cells loaded by reference to a poly(ADP-ribose) standard curve (0–25 pmol; Enzo Life Sciences).

Mouse liver samples were thawed and the wet weight was recorded prior to homogenization in 3 vol. of ice-cold iso-osmotic buffer (Ultra-Turrax T25; Janke and Kunkel). The homogenate was diluted with iso-osmotic buffer to yield a final dilution of 1:2000. The protein content was measured by the colorimetric Pierce protein assay (Thermo Scientific) prior to assaying PARP-1 activity as described above.

Statistical analysis

Each sample was analysed in triplicate (PARP activity) or in duplicate (PARP-1 expression) and results were expressed as the means. The normality of the data distribution was tested by Shapiro–Wilk and D'Agostino and Pearson tests (GraphPad). The distribution of PARP-1 activity and expression was highly skewed and so a log transformation was applied in order to obtain a more Gaussian-like distribution. Mean log-transformed PARP-1 activity and expression was compared between sexes and between HV and CP using the Student's t test. ANCOVA (analysis of co-variance) and linear regression were used to determine associations between activity and expression, and between activity, expression and age, weight and sex. The χ2 test and Freeman–Halton extension of the Fisher's exact test were used for analysis of genotype frequencies. The level of significance (P) was set at 0.05.

RESULTS

PARP-1 gene polymorphisms

2444T>C SNP (V762A) in the catalytic domain

The 2444T>C SNP (rs1136410), resulting in the amino acid substitution V762A in the PARP-1 catalytic domain is reported to lead to reduced PARP-1 catalytic activity. The genotype frequencies are given in Table 2. There was no evidence that the distribution was not in Hardy–Weinberg equilibrium in HV and CP (P=0.88 and P=0.31 respectively). The variant (minor) allele frequency (MAF) for both groups was 14% and was within the reported range for studied populations (5–33%) [2123]. There was no difference in genotype distribution between HV and CP (P=0.9). Neither the T/C nor the C/C genotypes were associated with an increased risk of cancer when compared with T/T [OR (odds ratio), 1.06; 95% CI (confidence interval), 0.50–2.20 and 0.48; 95% CI, 0.03–7.81 respectively].

View this table:
Table 2 Distribution of PARP-1 genotypes

(a) 2444T>C (V762A) polymorphism. For active site, 2444T>C SNP C/C is a variant/variant. P=0.9. (b) (CA)n polymorphism. Promoter polymorphism: SS, short alleles (CA)11–12/(CA)11–12; SL, short/long alleles (CA)11–12/(CA)13–20; LL, long/long alleles (CA)13–20/(CA)13–20. P=0.003. There was no statistically significant difference in genotype distribution between HV and CP.

(CA)n microsatellite instability in PARP-1 promoter region

It has been proposed that a long CA polymorphism in the promoter region of the PARP-1 gene may result in increased PARP-1 expression. We therefore investigated the length of these microsatellite repeats in all subjects' genomic DNA samples. Analysis of the allele frequencies in HV and CP revealed the presence of the two most common alleles, namely (CA)11 and (CA)15, which formed the three most common genotypes: HV, (CA)11/(CA)11 (61%), (CA)11/(CA)15 (16%) and (CA)15/(CA)15 (9%) and CP, (CA)11/(CA)11 (78%), (CA)11/(CA)15 (14%) and (CA)15/(CA)15 (3%). As previously established [24] we grouped the CA microsatellite into two alleles: short, comprising (CA)11–(CA)12 and long, comprising (CA)13–(CA)20. Genotype frequencies given by this biallelic approach are presented in Table 2. There was a statistically significant difference in the genotype distribution between CP and control subjects (P=0.003); CP had a higher frequency of the SS [short alleles (CA)11–12/(CA)11–12] genotype compared with HV (80% against 59%). The SS genotype was significantly associated with an increased risk of cancer (OR, 5.22; 95% CI, 1.79–15.24) using LL [long/long alleles (CA)13–20/(CA)13–20] as the reference group.

PARP-1 expression

We found that PARP-1 protein expression (see, for example, Supplementary Figure S1 at http://www.BiochemJ.org/bj/436/bj4360671add.htm), successfully analysed in 44 HV subjects, showed a large variation between the lowest and the highest expression in subjects [CV (coefficient of variation)=95%]; range 0.02–0.78 ng of PARP-1/μg protein with a mean value of 0.21 ng/μg and a median value of 0.12 ng/μg. Significant variation in PARP-1 expression was also observed in CP (0.03–1.55 ng/μg, CV=104%) with a mean value of 0.23 ng/μg and median value of 0.16 ng/μg. We did not observe any statistically significant difference in expression between HV and CP (P=0.18, Figure 1A) or men and women either in HV or in CP (P=0.1 and P=0.13 respectively, Figure 1B).

Figure 1 PARP-1 expression levels in HV and CP

(A) PARP-1 expression (log transformation) in PBMCs from HV (n=44) and CP (n=118). Each data point is a single individual and the horizontal line is the mean for the respective groups of samples. The difference between the two groups is not statistically significant (P=0.18 by Student's t test). (B) PARP-1 expression (log transformation) in male HV (n=12), female HV (n=32), male CP (n=65) and female CP (n=53). Each data point is a single individual and the horizontal line is the mean for the respective groups of samples. The difference between the two groups is not statistically significant (P=0.1 and P=0.13 for HV and CP respectively).

PARP activity

There was a large inter-individual variation in PARP activity (see, for example, Supplementary Figure S2 at http://www.BiochemJ.org/bj/436/bj4360671add.htm) in HV (n=56), with values ranging between 10 and 2190 pmol of PAR/106 PBMCs (CV=120%), mean value of 508.6 pmol/106 cells and median value of 260 pmol of PAR/106 cells. Similarly, we observed a large variation in PARP activity between CP (n=118, CV=137%) ranging between 10–2600 pmol/106 cells with mean value of 357.7 pmol/106 cells and median value of 160 pmol of PAR/106 cells. There was no statistically significant difference in PARP activity between HV and CP (P=0.45, Figure 2A). However, we observed a difference in PARP activity in HV between men and women (P=0.04, Figure 2B).

Figure 2 PARP activity in HV and CP

(A) PARP activity (log transformation) in PBMCs from HV (n=56) and CP (n=118). Each data point is a single individual and the horizontal line is the mean for the respective groups of samples. The difference between the two groups is not statistically significant (P=0.45 by Student's t test). (B) PARP activity (log transformation) in HV men (n=20), HV women (n=36), CP men (n=65) and CP women (n=53). Each data point is a single individual and the horizontal line is the mean for the respective groups of samples. The difference between HV men and women is statistically significant (P=0.04 by Student's t test).

Dependence of PARP activity on PARP-1 genotype and PARP-1 protein expression

We found no association between the 2444T>C genotype and PARP activity (T/T compared with T/C, P=0.24 and P=0.34, for HV and CP respectively, Figure 3A). It was not possible to perform any statistical analysis on the C/C genotype as we only had one sample with the C/C variant genotype in each group of subjects. We found no association between the (CA)n microsatellite polymorphism and PARP-1 expression (Figure 3B).

Figure 3 PARP activity in subjects with polymorphisms

(A) Scatter plot of PARP activity in HV and CP in subjects with 2444T>C T/T, T/C and C/C genotype. (B) Scatter plot of PARP-1 expression in HV and CP in subjects with promoter polymorphism: SS, short alleles (CA)11–12/(CA)11–12; SL, short/long alleles (CA)11–12/(CA)13–20; and LL, long/long alleles (CA)13–20/(CA)13–20.

We hypothesized that PARP activity would be dependent on the level of PARP-1 protein expression. There was a statistically significant but modest positive correlation between the level of PARP-1 protein expression and PARP activity in the HV (n=44, R2=0.19, P=0.003, Figure 4A). A positive but even weaker correlation was also found between PARP-1 expression and activity in CP (n=118, R2=0.06, P=0.01, Figure 4B).

Figure 4 Correlation between PARP-1 expression and PARP-1 activity

Correlation between PARP-1 expression (log transformation) and PARP-1 activity (log transformation) in HV (A) and in CP (B).

Demographic effects on PARP-1 expression and activity

In contrast with PARP-1 expression, comparison of gender differences in PARP-1 activity in HV (Figure 2B) revealed that men had significantly higher activity than women (P=0.04). The median PARP activity in PBMCs from men was 550 pmol of PAR/106 cells (range 10–1700 pmol of PAR/106 cells) with CV=83.7%. The median value for women was 130 pmol of PAR/106 cells (range 10–21900 pmol of PAR/106 cells) with CV=147.4%. The gender difference persisted after allowing for differences in PARP-1 expression (ANCOVA). On average, males had a 1.9-fold (95% CI, 1.2–2.8) higher level of PARP activity compared with females after adjusting for differences in PARP-1 expression (P=0.006).

Following the previous reports [7,16,17] showing that the age of the subject may affect PARP activity, we investigated the relationship between the age and weight of subjects and PARP-1 expression and activity. PARP activity was negatively correlated with age (P=0.02) in CP, but a similar association was not seen in HV (P=0.9) (Figure 5A). On average, for each 10 year increase in age, PARP activity in CP reduced by 19% (95% CI, 4–31%). PARP activity was not associated with weight in HV or CP (see Supplementary Figure S3 at http://www.BiochemJ.org/bj/436/bj4360671add.htm). We observed no association between PARP-1 expression and age (Figure 5B); however, PARP-1 expression was found to be significantly positively correlated with weight in CP (P=0.01). On average, for each 10 kg increase in weight, PARP-1 expression increased by 13% (95% CI, 3–23%).This association was not found in HV (see Supplementary Figure S4 at http://www.BiochemJ.org/bj/436/bj4360671add.htm).

Figure 5 Correlation between age and PARP activity/expression

(A) Correlation between age and PARP activity (log transformation). (B) Correlation between age and PARP-1 expression (log transformation).

In vivo studies of PARP activity in mice treated with oestrogen

To investigate the role of oestrogen in the regulation of PARP activity, we performed hormonal manipulation in mice. PARP activity in PBMCs was approximately 40% higher in male mice (920±20 pmol of PAR/106 cells) compared with female mice (570±70 pmol of PAR/106 cells, P=0.004) (Figure 6A). Castration led to a significant decrease in PARP activity (710±38 pmol of PAR/106 cells) (P=0.0005). Oestrogen supplementation of castrated male mice did not change the level of PARP activity. Paradoxically, oestrogen supplementation in female mice caused a significant increase in PARP activity in PBMCs (880±30 pmol of PAR/106 cells, P=0.003), bringing it to the level similar to that in control untreated male mice. In marked contrast, PARP activity was approximately 80% reduced in liver homogenates from oestrogen-treated female mice (0.15±0.05 pmol of PAR/mg of protein) compared with untreated female mice (0.9±0.42 pmol of PAR/mg of protein) (Figure 6B).

Figure 6 Effect of oestrogen on PARP activity

(A) Effect of oestrogen (E2) on the PARP activity in the CD1 males, castrated (castr.) males and female mice. Blood from three (controls) or five (treated) animals in each group was pooled and used as a single sample. Values are means±S.E.M. All P values were significant (Student's t test). (B) Effect of oestrogen treatment (E2) on the PARP activity in the liver homogenates from female mice. PARP activity was measured in triplicate per liver sample. Three livers per group were used.

PARP activity and patients' response to treatment

We evaluated patients' response to anticancer treatment in terms of toxicity in 44 patients. We have only chosen patients whose treatment was ‘PARP relevant’, that is, those agents known to cause more cytotoxicity and toxicity in PARP-null or inhibited cells or mice respectively. We studied patients treated with temozolomide and dacarbazine (alkylating agents), radiotherapy only (ionizing radiation), and radiotherapy in combination with the following chemotherapeutic agents: temozolomide, cisplatin (radiosensitizer, cross-linking agent), capecitabine (antimetabolite). Among the 44 patients studied, 15 (34%) developed toxicity grade 3 or greater (Table 3). The remaining 29 patients (66%) tolerated treatment well, with no toxicity or toxicity less than grade 3 (see Supplementary Table S1 at http://www.BiochemJ.org/bj/436/bj4360671add.htm). Patients treated with radiotherapy in combination with cisplatin were far more likely to experience high grade 3 or greater toxicity, (OR 13.2; 95% CI, 2.9–58.9). However, even after adjusting for exposure to cisplatin, there was no evidence of a relationship between high-grade toxicity and PARP activity or PARP-1 expression (P=0.5 and P=0.74, respectively), nor was there evidence of a relationship between high-grade toxicity and the 2444T>C genotype (P=0.22).

View this table:
Table 3 High grade toxicity (grade≥3) observed in patients treated with a PARP-relevant treatment

The genotype for the catalytic domain SNP (2444T>C) for each patient is shown. n=15. CAP, capecitabine; CIS, cisplatin; TMZ, temozolomide.

DISCUSSION

Since PARP activity may have profound implications for health, and as there is a wide inter-individual variation in PARP activity, the overall goal of the present study was to determine the mechanisms underlying this inter-individual variation. To this end, we measured polymorphisms in the PARP-1 gene that could affect its expression and activity, PARP-1 protein levels and demographic factors in relation to PARP activity in PBMCs from human subjects.

Our study revealed a large variation (CV=120% and 137% for HV and CP respectively) in PARP activity between individuals. Inter-individual variation in PARP-1 expression was much lower (CV=95% and 104% for HV and CP respectively). Although there was a positive correlation between PARP-1 expression and activity in the HV and CP, supporting the hypothesis that PARP activity reflects its abundance, the correlation was not as strong as expected, with only approximately 20% (HV) and <10% (CP) of the variation in activity explained by variation in expression.

In the 174 individuals we studied, we did not find that the 2444T>C SNP in the active site affected PARP-1 activity, which is similar to observations made by Cottet et al. [21] in 95 individuals. However, another study [4] of 354 individuals as well as in vitro studies [13,15] using purified PARP-1 enzyme did report a decrease in PARP-1 activity associated with the variant allele. Whether the difference is only detectable in large studies or whether the method used to determine PARP-1 activity (H2O2 stimulation [4] or in vitro analysis [13,15]) is responsible for the different findings is not possible to say at this time, but it is clearly worthy of further investigation using a standardized protocol. Similarly, the observed inter-individual variation in PARP-1 expression was not affected by the polymorphisms in the promoter region of PARP-1 gene; the two allelic approach, grouping all identified alleles into short and long and further correlation analysis with PARP-1 expression did not show any association between the level of PARP-1 protein and the length of microsatellite.

The lack of correlation between PARP-1 activity and genotype, with only a limited influence of PARP-1 expression, suggested that other factors may play a role in the regulation of PARP activity, such as demographics. We confirmed a previously reported [16] negative correlation between PARP activity and age. This was only seen in CP, where there was a 66-year difference between the youngest and oldest subject (22–88 years), although there was substantial overlap in the age distribution in the HV compared with the CP, the age distribution was narrower (41 years; 18–69) and hence a trend was more difficult to determine. For the first time, we showed that PARP activity was not associated with weight.

Our most striking and novel observation was a statistically significant difference (P=0.04) in PARP activity between men and women (HV). We found that younger women (<45 years, an age chosen based on epidemiological studies: http://www.nhs.uk/Conditions/Menopause) generally had lower activity (mean 414.1 pmol of PAR/106 cells, 95% CI, 133.1–695.1) than older women, who had intermediate activity (mean 427.5 pmol of PAR/106 cells, 95% CI, 73.6–781.3) between young women and men (mean 688.8 pmol of PAR/106 cells, 95% CI, 419.1–958.4), but the difference was not statistically significant. The gender differences found in HV were not found in CP (P=0.242), possibly because they were largely in the over 45 year-old group (94% of CP women). There are no previous reports of gender differences in PARP activity in humans, although sexual dimorphisms in PARP activity have been reported in animal models [25,26]. Our finding in humans and published data in animals suggesting hormonal regulation of PARP activity led us to conduct further investigations of hormone effects in mice. As with the human subjects, PARP activity in male mice was significantly higher than in females. PARP activity in castrated males was significantly reduced compared with intact males and similar to that in females. However, oestrogen supplementation failed to reduce PARP activity further in castrated males. Thus it seems more likely that gender differences in PARP activity in PBMCs are due to androgen-mediated stimulation rather than oestrogen-mediated inhibition of PARP activity. Consistent with this hypothesis is the observation that PAR formation was approximately 2-fold higher in brain tissue from male mice than females in a stroke model [27] and that PARP-1-mediated damage following cerebral ischaemia was significantly reduced in castrated mice [28]. However, oestrogen supplementation profoundly reduced PARP activity in the liver, consistent with the observed oestrogen-mediated PARP inhibition in the livers but not PBMCs of male mice treated with lipopolysaccharide [25]. This differential effect of oestrogen on different tissues warrants further investigation, as it may be relevant to diseases where PARP-1 activation has pathological effects (e.g. diabetes, [29]) and where there are sex differences in incidence or outcome (e.g. primary liver cancer [30]).

One of the aims of the present study was to correlate PARP activity with patients' response to treatment in terms of toxicity. The work presented here is the first study of PARP activity in relation to toxicity in patients receiving chemotherapy or radiotherapy. Patients vary in their response to chemotherapy and/or radiotherapy and the underlying mechanism comprises numerous different factors, including the patient's genetic profile (pharmacogenetics) and age. Previous studies demonstrate a potentiation of both anticancer activity and toxicity when cytotoxic substances are combined with PARP inhibitors [18,19,31]. These data suggest that PARP activity may not only have an impact on therapeutic response but also on the toxic response to chemotherapy or radiotherapy. We did not observe any significant difference in PARP activity between patients who suffered substantial toxicity (grade 3 and above) and those with no symptoms of toxicity. We confirmed the high levels of toxicity with cisplatin–radiotherapy combinations [32], but again the patients' PARP activity did not appear to influence the toxic response. Additionally, we observed that 40% of men, but only 21% of women, experienced high grade toxicity. In contrast with our findings, several clinical trials have reported greater toxicity in women [33,34].

Analysis of the CA repeat polymorphism and the two allelic approach, where alleles were grouped into short and long, revealed that the SS genotype was associated with increased risk of cancer (OR, 5.22; 95% CI, 1.79–15.24) compared with the LL genotype, which was under-represented in CP. To the best of our knowledge, this is the first time that the length of the CA microsatellite has been linked with cancer risk. Since we did not find an associated effect on PARP-1 expression or activity, the functional consequences of the short CA repeat that could explain the increased cancer risk remains to be determined.

In summary, in the present study we tried to find a possible explanation for the observed large inter-individual variation in PARP activity and if it had an impact on toxicity in patients, with a view to progression towards individualised therapy. We did not find any strong evidence that genetic factors play a role in determining PARP activity. The lack of any association between PARP-1 activity and genotype or sex in CP may indicate additional factors (e.g. stress hormones, interaction with other proteins and post-translational modifications) may play a role in PARP-1 activation. However, we did not find any association between PARP-1 activity and patients' response to treatment.

An unexpected observation was that PARP activity shows only a modest dependence on PARP-1 protein expression, indicating that endogenous or exogenous factors play a major role in regulating PARP activity. Importantly, we show for the first time a gender difference in PARP activity in the normal human population. Given that PARP activity can have implications for human health, further investigations of the role of hormones on PARP-1 activity and the tissue specificity of the effect are warranted.

AUTHOR CONTRIBUTION

Tomasz Zaremba, Sally Coulthard, Elizabeth Plummer and Nicola Curtin designed the experiments; Tomasz Zaremba and Elizabeth Plummer recruited patients and collected clinical data; Tomasz Zaremba conducted the experiments and analysed data; Michael Cole performed statistical analysis; Huw Thomas conducted in vivo experiments; and Tomasz Zaremba and Nicola Curtin wrote the manuscript.

FUNDING

T. Z. was supported by the Association for International Cancer Research (AICR) and N. J. C. is supported by Cancer Research U.K.

Acknowledgments

We thank Professor Alexander Bürkle for the anti-PAR 10H antibodies and Pfizer GRD for the AG014699 inhibitor. We also thank Rebecca Perret, Jane Margetts and staff from Newcastle General Hospital Ward 36 for help in design and conduct of the study and all the HV and CP who donated their blood.

Abbreviations: ANCOVA, analysis of co-variance; CI, confidence interval; CV, coefficient of variation; CP, cancer patients; HRP, horseradish peroxidase; HV, healthy volunteers; OR, odds ratio; PAR, poly(ADP-ribose); PARP, poly(ADP-ribose) polymerase; PBMC, peripheral blood mononuclear cell; SNP, single nucleotide polymorphism

References

View Abstract