Genetic polymorphisms in aquaporin 1 as risk factors for malignant mesothelioma and biomarkers of response to cisplatin treatment.

Background Malignant mesothelioma (MM) is an asbestos related aggressive tumor with poor prognosis. The aim of this study was to investigate if aquaporin 1 (AQP1) genetic polymorphisms influence the risk of MM and the response to cisplatin based MM treatment. Patients and methods The case-control study included 231 patients with MM and a control group of 316 healthy blood donors. All subjects were genotyped for three AQP1polymorphisms (rs1049305, rs1476597 and rs28362731). Logistic and Cox regression were used in statistical analysis. Results AQP1 rs1049305 polymorphism was significantly associated with MM risk in dominant model adjusted for gender and age (OR = 0.60, 95% CI = 0.37-0.96, Padj = 0.033). This polymorphism was also significantly associated with cisplatin based treatment related anaemia (unadjusted: OR = 0.49, 95% CI = 0.27-0.90, P = 0.021; adjusted: for CRP: OR = 0.52, 95% CI = 0.27-0.99, P = 0.046), with leukopenia (OR = 2.09, 95% CI = 1.00-4.35, P = 0.049) in dominant model and with thrombocytopenia (OR = 3.06, 95% CI = 1.01-9.28, P = 0.048) and alopecia (OR = 2.92, 95% CI = 1.00-8.46, P = 0.049) in additive model. AQP1 rs28362731 was significantly associated with thrombocytopenia (unadjusted: OR = 3.73, 95% CI = 1.00-13.84, P = 0.049; adjusted for pain: OR = 4.63, 95% CI = 1.13-19.05, P = 0.034) in additive model. Conclusions AQP1 may play a role in the risk of MM. Furthermore, AQP1 genotype information could improve the prediction of MM patients at increased risk for cisplatin toxicity.

Introduction

It is generally accepted that the risk of developing diseases and an individual’s response to the treatment may also depend on their genetic characteristics. In this study, we have focused on malignant mesothelioma (MM), which is a very aggressive cancer associated with the exposure to asbestos.1, 2, 3, 4 Most frequently it arises from pleura or peritoneum, but can also arise from other serous surfaces.5, 6

In Slovenia, the professional exposure to asbestos occurred mainly in asbestos cement industry, in construction, in manufacture of machinery and insulation materials, in maintenance of various means of transport, in textile industry and in other activities.7, 8, 9 Malignant mesothelioma is associated also with exposure to asbestos outside the workplace.5, 8, 9, 10 It is estimated that the incidence of MM will remain stable or will even increase in the near future due to the continuous presence of asbestos in buildings and to the long latent period after exposure to asbestos.11, 12 It is predicted that its incidence in the most industrialized countries will continue to increase until 2020,5 or even later.11

Due to an increasing incidence of MM and its poor prognosis, new prognostic and predictive biomarkers are needed.13 Symptoms of MM commonly occur only at late stages, therefore novel biomarkers for earlier diagnosis of MM and for establishing the response to treatment might be a promising opportunity for these patients.14 Several classes of potential biomarkers of MM have been studied so far, from serum peptides to genetic and epigenetic biomarkers, however with limited success. Among serum biomarkers, soluble peptides related to mesothelin (soluble mesothelin-related peptides, SMRP),15 fibulin-3,16 survivin13 have been studied, however none of them had sufficient predictive value as a standalone biomarker. It has been proposed, that biomarkers from two different molecular classes: protein and miRNA could be used in a combination to improve the biomarker sensitivity and specificity.14

Another approach was to investigate interin-dividual genetic variability in genes coding for key determinants of molecular pathogenesis of MM as potential biomarkers for prediction of the risk of MM as well as treatment response. Several studies have shown that polymorphisms in the genes involved in xenobiotic and oxidative metabolism or in DNA repair processes may play an important role in aetiology and pathogenesis of MM.17, 18, 19 The most commonly studied GSTM1 null polymorphism showed an increased risk for MM.17 Similarly, two variant alleles of XRCC1 and XRCC3 were associated with increased risk for MM.17 The carriers of at least one polymorphic NQO1 allele (CT and TT genotypes) had an increased risk of MM compared to those with CC genotype.19 A recent study showed also the association of FTO variability with MM susceptibility.20 On the other hand, MMP2 polymorphism was suggested to have a protective role in MM.21 Furthermore, two of the investigated MMP9 single nucleotide polymorphisms (SNPs) had significant but opposing effect on time to progression (TTP) and overall survival (OS) in MM.22 FASL-844 polymorphism could predict progression free survival (PFS) in MM patients receiving platinum based chemotherapy.23 Polymorphisms in REV1 and REV3L were also associated with the outcome of cisplatin based chemotherapy in MM.24 The results of the study showed that DNA repair gene polymorphisms XRCC1 may modify the response to gemcitabine-platinum combination chemotherapy in MM patients.18 Despite such high numbers of genetic factors investigated, the search for potential novel genetic biomarkers continues.

Aquaporins (AQPs) are small transmembrane proteins, which facilitate an osmotically controlled passage of water. Recent research indicated a key role of AQPs in human carcinogenesis.25, 26, 27 All key processes in cancer cells depend on water in the tumour microenvironment, therefore an enhanced transmembrane transmission of water is stimulated in comparison to normal cells. Overexpression of AQPs in the cell lines of the vascular endothelium and tumour cell lines suggests that AQPs may be closely related to the development and progression of a tumour.28 In some cancers AQP1 expression was also shown to participate in metastatic processes.29 In AQP1-knockout mice, xenograft tumour growth and angiogenesis were reduced, and significant necrosis occurred in the tumour tissues.30

The expression of AQP1 in MM tumour cells has been suggested to be an independent prognostic factor favouring survival in MM patients: higher levels of an AQP1 expression only in tumour cells, but not in vascular cells, predicted a better survival.31 Higher levels of AQP1 expression were also associated with a better course of the disease in MM, but with worse course of the disease in some other tumours such as breast cancer, melanoma, urothelial and pharyngeal carcinoma.32, 33, 34, 35 AQP1 is of interest as a potential biomarker in MM patients as it was shown to be an independent prognostic factor11 with high levels of its expression correlating with an increased survival.29, 31, 36 AQP1 expression also correlated with improved survival rates in MM with epithelioid component in comparison to AQP1-poor MM.37 Furthermore, AQP1 is also a possible new target for MM treatment,5 and there are already AQP1 blockers available which could be used for therapy.38

Genetic polymorphisms were reported in AQP1 gene, however according to our knowledge they have never been investigated in MM. A functional AQP1 rs1476597 (-783G/C) SNP leading to transcriptional activation of the AQP1 promoter and increased AQP1 mRNA expression in C allele carriers was associated with better survival in glioblastoma multiform patients with GG and GC genotype.39 Other AQP1 SNPs were studied in a variety of conditions, but not in cancer. Firm evidence suggested that AQP1 rs1049305 SNP could be involved in genetic susceptibility for development of water retention in patients with liver cirrhosis.40 The study in marathon runners reported a significant association between AQP1 rs1049305 and running performance. This study suggested that AQP1 rs1049305 polymorphism located in 3′-UTR, in interaction with miRNAs could influence the mRNA expression and AQP1 protein levels.41 Triathletes who carried AQP1 rs1049305 C allele had better running performance in comparison to GG genotype. This SNP was not associated with relative body weight change.41 It has been suggested that AQP1 rs10244884 could predict the risk of vaso-occlusion in sickle cell patients.42

The aim of the present study was to investigate the influence of AQP1 genetic polymorphisms on the risk of developing MM and response to cisplatin-based treatment.

Patients and methods

Study population

The case-control study included patients treated for mostly MM of pleura or also peritoneum at the Institute of Oncology Ljubljana from 2007 to the end of 2016. Control group consisted of blood donors from the Institute of Transfusion Medicine in Ljubljana and were over 40 years old.

The diagnosis of MM was made by means of thoracoscopy or video-assisted thoracoscopic surgery (VATS) in patients with pleural MM and by means of laparoscopy or laparotomy in peritoneal MM. The diagnosis was confirmed histopathologically by an experienced pathologist [15].15

Demographic and clinical data (age, gender, smoking, possible other diseases) from patients with MM were obtained from the medical records of the Institute of Oncology Ljubljana.

The following clinical indicators were used to evaluate the efficacy of treatment: response to treatment according to the modified criteria RECIST (Response Evaluation Criteria in Solid Tumours),43 PFS and overall survival (OS). The toxicity of the treatment was assessed according to NCI criteria (National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0).44

Ethical approval

The study was approved by the Republic of Slovenia National Medical Ethics Committee (41/02/09) and was carried out according to the Helsinki Declaration. All the subjects included in the study have signed the written informed consent.

Genotyping methods

DNA samples from 26 patients were isolated from peripheral venous blood with commercially available reagent sets (QIAamp DNA Mini Kit and Flexigene DNA Kit (Qiagen, Hilden, Germany)). For all other patients and controls DNA was already isolated from peripheral venous blood samples during the course of the previous studies.18, 45, 46, 47, 48

Based on the bioinformatics analysis, we selected the following SNPs: AQP1 rs1049305 G> C in 3’-untranslated region that may affect the binding of miRNA [41], AQP1 rs1476597 G> C in the 5’-regulatory region that may affect the binding of the transcription factors49 and AQP1 rs28362731 G> A that may affect splicing.

All the polymorphisms were genotyped using competitive allele specific PCR (KASPar) according to the manufacturer’s instructions (LGC Genomics, UK).

Statistical analysis

All statistical analyses were performed using IBM SPSS Statistics version 21.0 (IBM Corporation, Armonk, NY, USA). With the usual descriptive statistics we first described the characteristics of each variable separately. In order to assess the causal relationship between MM and the individual variables, we first used a univariate logistic regression. Both additive and dominant models were used to assess the effect of the selected AQP1 polymorphisms. Analysis was followed by the multivariate statistical modelling, taking into account the selected AQP1 polymorphisms and possible confounders such as age, gender, and smoking and significant clinical parameters. Hazard ratio (HR), 95% confidence interval (95% CI) and P-value were determined by Cox regression and median survival was determined by the Kaplan-Meier method.

In order to test the interactions between the selected AQP1 polymorphisms we introduced the logistic regression models with dummy variables.

Results

The clinical characteristics of MM patients are shown in Table 1. Among all 231 patients whose median (25%–75% range) age was 66 (58–73) years, men represented 73.6%. Epithelioid MM was present in 72.3% of patients. ECOG performance status 1 (48.1%) and 2 (39.0%) prevailed. Exposure to asbestos was confirmed in 73.8% of patients. Among all patients, 46.7% were smokers. In total 194 patients were treated with cisplatin based therapy.

Table 1

Description of all malignant mesothelioma (MM) patients (N = 231) and MM patients treated with cisplatin based chemotherapy (N = 194)

		All MM patients	MM patients cisplatin treated based with chemotherapy
Characteristic	Characteristic type	N (%)	N (%)
Age	Median (25%–75%)	66 (58–73)	65 (58–71.3)
Gender	Men	170 (73.6)	146 (75.3)
	Women	61 (26.4)	48 (24.7)
	I	18 (7.8)	15 (7.7)
	II	57 (24.7)	48 (24.7)
MM stage	III IV	69 (29.9) 66 (28.6)	62 (32.0) 50 (25.8)
	Peritoneal MM	20 (8.7)	18 (9.3)
	Undefined	1 (0.4)	1 (0.5)
	Epithelioid	167 (72.3)	147 (75.8)
Histological	Biphasic	26 (11.3)	21 (10.8)
type	Sarcomatoid	24 (10.4)	21 (10.8)
	Undefined	14 (6.0)	5 (2.6)
	0	15 (6.5)	15 (7.7)
ECOG	1	111 (48.1)	100 (51.5)
performance status	2	90 (39.0)	76 (39.2)
	3	15 (6.5)	3 (1.5)
Exposure to	No	59 (26.6)a	45 (23.3)c
asbestos	Yes	166 (73.8)	148 (76.7)
	No	120 (53.3)a	101(52.6)d
Smoking	Yes	105 (46.7)	91(47.4)
	Gemcitabine/ Cisplatin	132 (60.0)b	132 (68.0)
	Pemetrexed/Cisplatin	62 (28.2)	62 (32.0)
Treatment	chemotherapy Without	16 (7.3)	-
	Other forms of chemotherapy	10 (4.5)	-

Data are missing for: a 6 patients, b 11 patients, c 1 patient and d 2 patients. ECOG = Eastern Cooperative Oncology Group

In addition, 316 healthy blood donors, 235 men and 81 women, whose median (25%–75% range) age was 49 (45-55) years were also included in the molecular-genetic part of the study.

The genotype frequency distribution for the investigated polymorphisms in 231 MM patients and in 316 controls, their minor allele frequencies (MAF) and the risk of developing MM are shown in Table 2. The genotypes’ distribution was in Hardy-Weinberg equilibrium (HWE), except for the distribution of AQP1 rs1476597 in MM patients and also in healthy controls that were not consistent with HWE and therefore we excluded this polymorphism from further statistical analysis.

Table 2

Distribution of AQP1 genotypes in MM patients and controls and risk of MM

			Patients			Controls
SNP	Role	Genotype	N (%)	MAF	PHWE	N (%)	MAF	PHWE	OR (95% CI)	P	OR (95% CI)adj	Padj
rs28362731a	p.Gly165Asp	GG	210 (92.1)	0.039	0.535	288 (91.7)	0.041	0.444	Ref.	-	Ref.	-
		GA	18 (7.9)			26 (8.3)			0.95	0.871	0.94	0.885
									(0.51–1.78)		(0.38–2.30)
rs1049305b	c.*578G>C	GG	107 (46.5)	0.337	0.082	128 (40.8)	0.373	0.288	Ref.	-	Ref.	-
		GC	91 (39.6)			138 (43.9)			0.79	0.207	0.59	0.039
									(0.55–1.14)		(0.35–0.97)
		CC	32 (13.9)			48 (15.3)			0.80	0.390	0.63	0.199
									(0.48–1.13)		(0.32–1.27)
		GC+CC	123 (53.5)			186 (59.2)			0.79	0.181	0.60	0.033
									(0.56–1.12)		(0.37–0.96)
rs1476597c	c.-783G>C	GG	157 (68.0)	0.255	<0.001	220 (70.1)	0.247	<0.001
		GC	30 (13.0)			33 (10.5)
		CC	44 (19.0)			61 (19.4)

Data are missing for: a2 controls and 3 patients, b2 controls and 1 patient, c2 controls. adj = adjusted by gender and age; MAF = minor allele frequency; PHWE = P for Hardy-Weinberg equilibrium; Ref. = reference genotype

In univariate analysis no polymorphism was associated with the risk of developing MM (Table 2). Higher age was associated with a higher risk of developing MM (OR = 1.21, 95% CI = 1.17–1.25, P < 0.001) but gender (OR = 1.04, 95% CI = 0.71–1.53, P = 0.838) was not. AQP1 rs1049305 polymorphism was significantly associated with the risk of developing MM when adjusted for age and gender (OR = 0.59, 95% CI = 0.35–0.97, Padj = 0.039 in additive model; OR = 0.60, 95% CI = 0.37–0.96, Padj = 0.033 in dominant model). AQP1 rs28362731 was not significantly associated with the risk of developing MM even when adjusted for age and gender (Table 2).

Clinical characteristics of MM patients treated with cisplatin based chemotherapy are presented in Table 3. The majority (68.0%) of patients were treated with gemcitabine in combination with cisplatin. In chemotherapy response a third (32.8%) of patients responded with partial response (PR) and only in few patients (3.2%) the response was complete (CR). A half (49.5%) of patients had stable disease (SD) and a few (14.5%) of them had progressive disease (PD). Median progression free survival (PFS) was 7.8 months, median overall survival (OS) 18.1 months and median follow-up from the start of chemotherapy 49.2 months.

Table 3

Clinical characteristics of MM patients treated with cisplatin based chemotherapy (N = 194)

Characteristic		N (%)
	Gemcitabine and cisplatin	132 (68.0)
Chemotherapy type	Pemetrexed and cisplatin	62 (32.0)
	Complete response (CR)	6 (3.2)
	Partial response (PR)	61 (32.8)
Chemotherapy responsea	Stable disease (SD)	92 (49.5)
	Progressive disease (PD)	27 (14.5)
Progression of diseaseb	No	20 (10.5)
	Yes	171 (89.5)
	No	58 (29.9)
Death	Yes	136 (70.1)
PFS	Median (25%–75%) (month)	7.8 (5.3–13.8)
OS	Median (25%–75%) (month)	18.1 (9.4–28.7)
Follow-chemotherapy up from the start of	Median (25%–75%) (month)	49.2 (18.9–75.5)
CRP	Median (25%–75%)	20.5 (9–58)
LDH	Median (25%–75%)	2.67 (2.26–3.11)
Painb	No	79 (41.4)
	Yes	112 (58.6)
	No	68 (35.8)
Weight lossc	Yes	122 (64.2)

Data are missing for: a8 patients, b3 patients, c4 patients. CRP = C reactive protein; LDH = lactate dehydrogenase; OS = overall survival; PFS = progression free survival

In the survival analysis, AQP1 rs28362731 and AQP1 rs1049305 were not significantly associated with PFS or with OS when patients were treated with cisplatin based chemotherapy (Table 4). Even when adjusted for histological type of MM, smoking, weight loss and CRP, AQP1 polymorphisms were not significantly associated with PFS. Likewise associations with OS remained insignificant after adjustment for the histological type of MM, smoking and CRP (data not shown). In the chemotherapy response, AQP1 rs28362731 and AQP1 rs1049305 were not significantly associated with response rate when patients were treated with cisplatin in combination with either gemcitabine or pemetrexed (Table 4). These associations remained insignificant when adjusted for loss of weight and CRP (data not shown).

Table 4

Influence of AQP1 SNP on survival and chemotherapy response in MM patients

		Progress free	survival		Overall survival			Chemotherapy response
SNP	Genotype	PFS median (25%–75%) month	HR (95% CI)	P	OS median (25%–75%) month	HR (95% CI)	P	Poor response N (%)	Good response N (%)	OR (95% CI)	P
	GG	7.7 (5.2–13.6)	Ref.	-	18.1 (9.1–28.0)	Ref.	-	112 (65.1)	60 (34.9)	Ref.	-
rs28362731	GA	11.1 (7.0–14.7)	0.72 (0.39–1.33)	0.299	26.5 (14.4–47.8)	0.56 (0.26–1.19)	0.130	6 (54.5)	5 (45.5)	1.56 (0.46–5.31)	0.481
	GG	7.9 (5.4–12.1)	Ref.	-	18.1 (9.0–26.8)	Ref.	-	55 (64.7)	30 (35.3)	Ref.	-
	GC	7.8 (5.2–15.0)	0.80 (0.58–1.11)	0.187	22.1 (10.1–29.7)	0.72 (0.50–1.05)	0.091	43 (58.1)	31 (41.9)	1.32 (0.70–2.51)	0.394
rs1049305	CC	7.4 (4.8–14.1)	0.92 (0.59–1.46)	0.736	13.3 (8.1–25.4)	1.10 (0.67–1.80)	0.712	20 (76.9)	6 (23.1)	0.55 (0.20–1.52)	0.248
	GC+CC	7.8 (4.9–15.0)	0.83 (0.62–1.13)	0.233	18.2 (9.5–28.7)	0.81 (0.58–1.14)	0.220	63 (63.0)	37 (37.0)	1.08 (0.59–1.97)	0.810

SNP = single nucleotide polymorphisms; OS = overall survival; PFS = progression free survival; Ref. = reference genotype

The association between SNPs and side effects in cisplatin based treatment is shown in Tables 5 and 6. AQP1 rs1049305 was significantly associated with anemia grade ≥ 2 both in additive and dominant genetic model (additive model for genotype GC: OR = 0.40, 95% CI = 0.20–0.78, P = 0.007; dominant model OR = 0.49, 95% CI = 0.27–0.90, P = 0.021). The associations remained significant also when adjusted for CRP (OR = 0.46, 95% CI = 0.23– 0.92, P = 0.029 in additive model; OR = 0.52, 95% CI= 0.27–0.99, P = 0.046 in dominant model). AQP1 rs1049305 was also significantly associated with thrombocytopenia in additive model for genotype CC (OR = 3.06, 95% CI = 1.01–9.28, P = 0.048), but not in dominant model. AQP1 rs1049305 was also significantly associated with the risk of leukopenia (additive model for genotype CC: OR = 3.03, 95% CI = 1.10–8.38, P = 0.033; dominant model OR = 2.09, 95% CI = 1.00–4.35, P = 0.049). Furthermore, there was a significant association of AQP1 rs1049305 with alopecia in additive model for genotype CC (OR = 2.92, 95% CI = 1.00–8.46, P = 0.049), however, this SNP was not associated with neutropenia, nephrotoxicity or nausea and/or vomiting.

Table 5

Association between AQP1 SNPs and haematological side effects of cisplatin based treatment (N = 176)

				Anemia grade ≥ 2a					Thrombocytopeniab			Leukopenia grade ≥ 2c		Neutropenia grade ≥ 2
SNP	Genotype	N (%)		OR (95% CI)	P	OR (95% CI)adj1	P adj1	N (%)	OR P (95% CI)	OR (95% CI)adj2	Padj2	N OR (%) (95% CI)	P	N (%)	OR P (95% CI)
rs28362731	GG	79 (49.4)		Ref.	-	Ref.	-	21 (13.3)	Ref. -	Ref.	-	39 Ref. (25.2)	-	59 (36.4)	Ref. -
	GA	3 (27.3)		(0.100.38 –1.50)	0.169	(0.130.53 –2.12)	0.370	(36.44 )	(1.003.73 –13.84) 0.049	(1.134.63 –19.05)	0.034	(18.22 ) (0.140.66 –3.19)	0.606	3 (27.3)	(0.170.66 –2.56) 0.543
	GG	(56.846 )		Ref.	-	Ref.	-	(12.510 )	Ref. -	Ref.	-	(14 17.5) Ref.	-	25 (30.9)	Ref. -
	GC	(34.323 )		(0.200.40 –0.78)	0.007	(0.230.46 –0.92)	0.029	(11.88 )	(0.350.93 –2.52) 0.892	(0.240.71 –1.08)	0.529	(18 27.7) (0.821.81 –3.99)	0.144	26 (37.1)	(0.671.32 –2.60) 0.416
rs1049305	CC	(52.013 )		(0.340.82 –2.03)	0.674	(0.280.74 –1.94)	0.536	(30.47 )	(1.013.06 –9.28) 0.048	(0.692.18 –6.94)	0.185	(36.19 ) (1.103.03 –8.38)	0.033	11 (45.8)	(0.751.90 –4.81) 0.178
	GC +CC	(39.136 )		(0.270.49 –0.90)	0.021	(0.270.52 –0.99)	0.046	(16.515 )	(0.581.38 –3.28) 0.463	(0.431.07 –2.69)	0.885	(30.727 ) (1.002.09 –4.35)	0.049	37 (39.4)	(0.781.45 –2.72) 0.242

Data are missing for: a2 patients, b4 patients, c7 patients. adj1 = adjusted by CRP; adj2 = adjusted by pain at diagnosis; SNP = single nucleotide polymorphisms

Table 6

Associations between AQP1 SNPs and non-haematological side effects of cisplatin based treatment (N = 176)

			Alopeciaa			Nephrotoxicityb			Nausea/Vomitingc
SNP	Genotype	N (%)	OR (95% CI)	P	N (%)	OR (95% CI)	P	N (%)	OR(95% CI)	P
	GG	60 (45.8)	Ref.	-	74 (46.8)	Ref.	-	73 (53.7)	Ref.	-
rs28362731	GA	5 (55.6)	1.48 (0.38–5.76)	0.572	3 (27.3)	0.43 (0.11–1.66)	0.219	5 (55.6)	1.08 (0.28–4.19)	0.913
	GG	30 (46.2)	Ref.	-	35 (43.8)	Ref.	-	36 (52.9)	Ref.	-
	GC	20 (35.7)	0.65 (0.31–1.35)	0.246	34 (50.0)	1.29 (0.67–2.46)	0.448	26 (44.8)	0.72 (0.36–1.46)	0.364
rs1049305	CC	15 (71.4)	2.92 (1.00–8.46)	0.049	10 (43.5)	0.99 (0.39–2.52)	0.982	15 (71.4)	2.22 (0.77–6.41)	0.140
	GC+CC	35 (45.5)	0.97 (0.50–1.89)	0.934	44 (48.4)	1.20 (0.66–2.20)	0.547	41 (51.9)	0.96 (050–1.84)	0.900

Data are missing for: a 33 patients, b 4 patients and c 28 patients. SNP = single nucleotide polymorphisms

AQP1 rs28362731 GA genotype was significantly associated with thrombocytopenia (OR = 3.73, 95% CI = 1.00–13.84, P = 0.049). This association remained significant when adjusted for pain at diagnosis (OR = 4.63, 95% CI = 1.13–19.05, P = 0.034).

The investigated polymorphisms did not statistically significantly influence neutropenia grade ≥ 2, nephrotoxicity or nausea and/or vomiting (Tables 5 and 6).

Multiplicative interaction analysis did not show any interactions between AQP1 rs28362731 and AQP1 rs1049305 polymorphisms and the risk of developing MM (OR = 1.22, 95% CI = 0.33–4.56, P = 0.771). Additionally, interactions between rs28362731 and rs1049305, rs28362731 and smoking and rs1049305 and smoking did not significantly influence the risk of occurrence of side effects (Table 7).

Table 7

Influence of interactions on the risk of occurrence of side effects

	Interaction 1rs28362731		Interaction 2		Interaction 3
Side effect	-rs1049305	P1	rs28362731 - smoking	P2	rs1049305 - smoking	P3
	OR (95% CI)		OR (95% CI)		OR (95% CI)
Anemia grade ≥ 2a	1.84 (0.10–32.37)	0.676	-	0.999	0.34 (0.10–1.16)	0.085
Leukopenia grade ≥ 2b	0.95 (0.04–23.07)	0.974	-	0.999	0.92 (0.21–4.02)	0.915
Neutropenia grade ≥ 2	0.55 (0.03–9.76)	0.686	7.55 (0.39–145.1)	0.180	0.67 (0.19–2.35)	0.526
Thrombocytopeniac	1.73 (0.11–26.38)	0.693	3.06 (0.20–46.56)	0.422	0.95 (0.16–5.66)	0.955
Nephrotoxicityc	0.68 (0.04–11.98)	0.794	-	0.999	1.01 (0.30–3.43)	0.982
Alopeciad	2.06 (0.11–40.01)	0.633	-	0.999	0.60 (0.16–2.29)	0.453
Nausea/Vomitinge	2.12 (0.11–40.98)	0.620	6.83 (0.35–132.4)	0.204	0.71 (0.19–2.64)	0.608

Data are missing for: a 2 patients, b 7 patients, c 4 patients, d 33 patients and e 28 patients. Interaction 1: interaction between rs28362731 and rs1049305. Interaction 2: interaction between rs28362731 and smoking. Interaction 3: interaction between rs1049305 and smoking.

Haplotypes AQP1 GG, GC and AG (5’ ➝ 3’: rs28362731, rs1049305) were not significantly associated with the risk of developing MM even when adjusted for gender and age (Supplementary table).

Discussion

In the present study we investigated the influence of AQP1 genetic polymorphisms on the risk of developing MM as well as the associations with response to cisplatin based treatment. The important novel finding of our study is that the AQP1 genetic variability might contribute to the risk of developing MM. Furthermore, we have shown the associations with the development of side effects of cisplatin based treatment.

AQP1 rs1049305 polymorphism was significantly associated with the risk of developing MM, but only after adjustment for gender and age. AQP1 rs1049305 GC heterozygotes had significantly lower risk of developing MM in the additive model, as well as the carriers of at least one polymorphic C allele in the dominant model in comparison to GG wild type. This polymorphism is located in the 3’-untranslated region, therefore it could affect the binding of miRNA and AQP1 expression levels, however, the functionality of this polymorphism remains to be determined. On the other hand, AQP1 rs28362731 was not significantly associated with the risk of developing MM in our study.

The statistical analyses have shown that the genotype distribution for the third investigated polymorphism AQP1 rs1476597 was not in accordance with HWE, so it had to be excluded from further analysis. In this polymorphism, the substitution of G for C was associated both with increased transcriptional-activation of the AQP1-promoter and with increased AQP1 mRNA expression.39 This is the only AQP1 polymorphism that has been investigated in cancer so far and was associated with survival-time in glioblastoma multiforme patients.39 This study used a pyrosequencing approach and reported that genotype distribution for AQP1 rs1476597 was in accordance with HWE.39 We have checked that there was no genotyping error, so deviation from HWE could be interpreted either as a potential influence of the fact, that this polymorphism may be triallelic (G/C/A) or that the polymorphisms in the proximity could affect the binding of our primers from the reaction mixture.

We have also assessed the impact of AQP1 haplotypes, but they were not significantly associated with the risk of developing MM, not even when adjusted for age and gender.

Our study also showed that AQP1 rs1049305 and AQP1 rs28362731 were not significantly associated with PFS, OS or response rate when patients were treated with cisplatin in combination with either gemcitabine or pemetrexed. However, it has been suggested that AQP1 may be an independent prognostic factor in MM, and that higher expression of AQP1 in tumor cells, but not in vascular cells was significantly associated with better survival.31 It has also been shown that AQP1 expression significantly influenced the course of MM, regardless of the therapy or prognostic factors including histologic subtype, pathologic stage, gender, and age at time of diagnosis.31

We have also observed that AQP1 rs1049305 was significantly associated with some of the treatment side effects such as anemia, leukopenia, thrombocytopenia and alopecia, but not with neutropenia, nephrotoxicity or nausea and/or vomiting. On the other hand, AQP1 rs28362731 was significantly associated only with thrombocytopenia. Multiplicative interaction analysis did not show any interaction between AQP1 rs28362731 and AQP1 rs1049305 polymorphisms and the risk of occurrence of treatment related side effects. Similarly, side effects were not influenced by interactions between either of the studied polymorphism and smoking.

The major limitation of our study was that we had no information on asbestos exposure in healthy controls. Furthermore, MM patients were older than controls, as blood donors can only be up to 65 years old, however we accounted for that with adjustment for age in the statistical analysis. Despite the limited number of patients included in our study, all patients were monitored in the same institution and by the same oncologists, so there were no differences in the clinical assessments. Furthermore, all the patients and controls came from an ethnically homogeneous Slovenian population, so there were no differences due to genetic heterogeneity.50, 51

Our study brings novel findings of the associations between AQP1 genetic variability and the risk of developing MM that has not been previously investigated. Furthermore, it shows the impact of AQP1 polymorphisms on the development of cisplatin treatment related side effects. It needs to be determined if the addition of these polymorphisms to previously described clinical-pharmacogenetics models could improve the prediction of treatment related side effects in MM patients [48]. Better understanding of pharmacogenetic polymorphisms would allow an individualized approach and better outcomes of cisplatin treatment in patients with MM.

In conclusion, our study suggests that the investigated AQP1 polymorphisms may contribute to the risk of developing MM and cisplatin treatment related side effects, however our findings need to be validated in independent MM patient cohorts and in other cancers.