Phenylethanolamine N-methyltransferase gene polymorphisms associate with crisis pain in sickle cell disease patients

Abstract

Aim:

Phenylethanolamine N-methyltransferase (PNMT) catalyzes the conversion of sympathetic neurotransmitter norepinephrine to epinephrine. We examined the association of PNMT polymorphisms with acute and chronic pain in sickle cell disease (SCD).

Methods:

Utilization of emergency care owing to painful crisis was used as a marker for acute pain in 131 patients with SCD.

Results:

rs876493 A allele, rs2934965 T allele and rs2941523 G allele were significantly associated with decreased utilization (p ≤ 0.05). rs876493 A allele showed association with utilization in females (p = 0.003), not males (p = 0.803). rs2934965 T allele and rs2941523 G allele were predicted to cause loss of putative transcription factor binding sites. This is the first report of the association of PNMT polymorphisms with acute crisis pain in SCD. Together with our previous findings in catechol-o-methyltransferase, polymorphisms in catecholamine metabolizing enzymes appear to primarily influence acute pain in SCD.

Sickle cell disease (SCD) is a fatal blood disorder caused by the inheritance of mutations in the gene coding for the hemoglobin β-chain subunit [1]. The more severe mutation (HbS) favors the formation of extended polymeric tendrils of hemoglobin in the deoxygenated state precipitating the characteristic sickle shape red blood cells (RBCs). Short-lived, with decreased oxygen carrying capacity, the sickled RBCs often undergo intravascular hemolysis due to their rigidity and instability [2,3].

Consequently, sickled RBCs cluster thereby restricting blood flow and imparting endothelial dysfunction [1,4,5]. Poor gas exchange provokes an ischemic-like condition, commonly referred to as vaso-occlusive crisis (VOC), characterized by severe acute pain and target organ damage [1,6,7]. Acute pain presents itself during VOC and its etiology has been extensively studied; however, chronic pain in SCD is little understood. One study estimated that a third of all adult SCD patients experience chronic noncrisis pain almost daily [8]. While another study found that 80% of adult outpatients with SCD reported to be experiencing a constant pain pattern and >90% of them selected neuropathic pain descriptors to describe their pain quality [9]. A more recent study reported that on an average >50% of SCD patients aged 20 or older experience chronic pain [10]. Although the exact mechanisms are unknown, a number of working models have been proposed, including a state of heightened sensitivity secondary to exhaustive pain signaling, elevations in reactive oxygen species [6,7], peripheral nerve damage and central nerve sensitization attributed to the manifestation of allodynia and hyperalgesia [11,12]. While both opioid and non-opioid-based drugs are employed to manage acute and chronic sickle cell pain, the wide variability in the severity, frequency and pattern of the pain phenotypes makes pain management a challenge [13,14].

Phenylethanolamine N-methyltransferase (PNMT) is the enzyme responsible for the methylation of norepinephrine into epinephrine in the adrenal medulla [15–19]. Norepinephrine is a known endogenous pain modulator with predominant inhibitory effects [6]. On the other hand, epinephrine can sensitize nociceptive neurons resulting in hyperalgesia [20–22]. Moreover, studies in murine models have shown that epinephrine can mediate vaso-occlusion by increasing sickle RBC adhesion to the endothelium [23]. Thus, we hypothesized that genetic variants of the PNMT gene that alter its expression or activity may contribute to the heterogeneity of sickle cell pain by modulating levels of norepinephrine and epinephrine in the body.

The PNMT gene is composed of three exons interspaced by two intronic regions, and preceded by a promoter region [24,25]. Studies have shown that PNMT expression is regulated by a number of transcription factors located at the promoter region [26–28] and polymorphisms of PNMT have been implicated in altered transcriptional activity [16,25]. To our knowledge, polymorphisms of the PNMT gene have not been previously studied in the context of pain in SCD. In this study, we studied polymorphisms of the PNMT gene and their associations with acute and chronic pain in SCD.

Methods

Subject recruitment

Subjects were recruited during their scheduled visits at the University of Illinois Hospital and Health Sciences System in Chicago, Illinois. During these visits, blood or buccal swabs were collected as samples for subsequent DNA extraction and genotyping. Written informed consent was obtained from each subject and the study was approved by the University of Illinois at Chicago’s Institutional Review Board.

DNA extraction & genotyping

DNA extraction technique varied depending on the sample type. For buccal swabs, a modified phenol/chloroform protocol was used to isolate DNA [29]. Blood samples were either subjected to a modified salting-out procedure [30] or processed using the QuickGene-mini80 isolation device (AutoGen, MA, USA). Post DNA extraction, purified samples were genotyped utilizing the MassARRAY iPLEX Platform (Sequenom, CA, USA). Single-nucleotide polymorphisms (SNPs) included in the study had >95% genotyping call rate and were in Hardy–Weinberg equilibrium as determined by a χ² goodness-of-fit test (p > 0.05).

Pain assessment

The marker for baseline chronic pain was the composite pain index (CPI) score. Subjects rated their noncrisis chronic pain symptoms on a computerized reporting tool that employs the McGill Pain Questionnaire, as described in previous studies [31,32]. These inputs were weighted and aggregated to obtain CPI scores.

The marker for acute crisis pain was acute care utilization, that is, number of visits to the emergency department or acute care center prompted by crisis pain. A history of each subject’s utilization was enumerated over a period of 12 months from their baseline chronic pain assessment. Data were collected via medical records at the emergency department, University of Illinois Health System and phone calls to enrolled subjects.

While the McGill Pain Questionnaire is a well-established method of pain scoring and has been validated in sickle cell cohorts [32], it is important to note that all pain-assessment methods and tools are limited by the inherent subjective quality of pain as a symptom.

Statistics & bioinformatics

Statistical analysis was performed using R v3.4.0 (R Foundation for Statistical Computing, Vienna, Austria). For association analysis, additive, dominant, and recessive genetic models were utilized [33,34]. The major allele served as the reference allele and the minor allele as the risk allele in all models. In the additive model, effect of each minor allele is determined. The dominant model estimates the effect of carrying one or both minor alleles, and the recessive model estimates the effect of being homozygous with the minor alleles. CPI scores were modeled using multiple linear regression, and utilization scores were modeled using negative binomial regression. Models were adjusted for covariates: age, sex and sickle cell type. Sex-specific analysis was performed in only the dominant model owing to limited sample size in each sub-category. Linkage disequilibrium plots were generated using the software, Haploview v4.2 (Broad Institute, MA, USA). Identified haplotypes were analyzed for association with CPI and utilization groups. The analysis of utilization groups involved comparison of subjects without utilization versus those with utilization and also between subjects with low (zero to three events) utilization versus those with high (four or more events) utilization. Bioinformatics tools SNPinfo:FuncPred (https://snpinfo.niehs.nih.gov/snpinfo/snpfunc.html) and PROMO (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3) were used to predict if SNPs were in or near putative transcription factor binding sites (TFBS) and potential functional roles [35–37]. Input sequences for PROMO were obtained from the dbSNP database.

Results

Blood samples collected from a cohort of African-American subjects were used to genotype four SNPs on the PNMT gene. A summary of the baseline characteristics of the subjects is shown in Table 1. In total, 131 subjects were included of whom 77.9% were homozygous with sickle hemoglobin. The average subject age was 34 years (34.3 ± 11.8) with females accounting for 66% of the study population. As a group, subjects of this cohort had an average utilization number of 4.5 ± 5.3 with the numbers ranging from 0 to 38; and an average CPI of 40.6 with a range of 14.8–86.5 on a scale of 0–100, highlighting the wide variability in pain phenotypes, both acute and chronic pain.

Table 1. . Demographic characteristics (n = 131).

Demographics	Values	Phenotype	Values
Age (years)	Mean ± SD: 34.3 ± 11.8	Composite Pain Index (0–100)	Mean ± SD: 40.6 ± 13.4
	Range: 15–70		Range: 14.8–86.5
Sex, n (%)	Male: 45 (34.4)	Utilization score	Mean ± SD: 4.5 ± 5.3
	Female: 86 (65.6)		Range: 0–38
Sickle cell type† n (%)	SCD-SS: 102 (77.9)	Utilization groups n (%)	Zero (0): 19 (14.5)
	SCD-SC: 15 (11.5)		Low (1–3): 56 (42.7)
	Other: 14 (10.7)		High (≥4): 56 (42.7)

^†Sickle cell types: SCD-SS, SCD-SC, others include SCD-Sβ+ and SCD-Sβ°.

SCD: Sickle cell disease; SCD-Sβ°: Sickle cell disease-sickle β° thalassemia; SCD-Sβ+: Sickle cell disease-sickle β+ thalassemia; SCD-SC: Sickle cell disease-sickle hemoglobin C; SCD-SS: Sickle cell disease-homozygous hemoglobin S; SD: Standard deviation.

The major and minor allele relationships for the PNMT polymorphisms as well as their genotype frequencies within the recruited population are shown in Table 2. None of the four SNPs showed significant deviation from Hardy–Weinberg equilibrium. Additive, dominant and recessive genotype models were applied to discern the minor allele’s pattern of penetrance, if any, for each of the four PNMT polymorphisms. Analyses with CPI scores did not yield any significant associations (Table 3). In the analysis of utilization scores, three of the four SNPs (rs2934965, rs876493, rs2941523) achieved a statistically significant reduction in the utilization rate in at least one genetic model (Table 4). The T allele of rs2934965[G>T] exhibited a significant reduction in utilization rate in the additive model (incidence risk ratio = 0.61; p = 0.044). The allele A of rs876493[G>A] was found to be significantly associated with a reduction in the utilization rate in both the additive and dominant models (additive: B = 0.68; p = 0.012; dominant: B = 0.67; p = 0.030). The allele G of rs2941523[C>G] achieved statistically significant reduction in utilization rate in both the additive and recessive models (additive: B = 0.68; p = 0.033; recessive: B = 0.12; p = 0.017). In contrast, the synonymous variant rs5638 was not found to be associated with a significant change in utilization rate.

Table 2. . Phenylethanolamine N-methyltransferase polymorphism genotypes and allele frequencies.

dbSNP ID	Location	Allele	N (%)	Genotype	N (%)
rs5638	GRCh38 chr17: 39669996	A	217 (84.1)	AA	89 (69.0)
		G	41 (15.9)	AG	39 (30.2)
				GG	1 (0.8)
rs2934965	GRCh38 chr17: 39667402	G	241 (92.0)	GG	111 (84.7)
		T	21 (8.0)	GT	19 (14.5)
				TT	1 (0.8)
rs876493	GRCh38 chr17: 39668292	G	173 (69.2)	GG	57 (45.6)
		A	77 (30.8)	GA	59 (47.2)
				AA	9 (7.2)
rs2941523	GRCh38 chr17: 39667508	C	222 (84.7)	CC	94 (71.8)
		G	40 (15.3)	CG	34 (26.0)
				GG	3 (2.3)

Table 3. . Association analyses of Composite Pain Index with phenylethanolamine N-methyltransferase polymorphisms.

Single-nucleotide polymorphism	Model	B† (97.5% CI)	p-value
	Additive (AA vs AG vs GG)	1.51 (-3.42–6.43)	0.546
rs5638	Dominant (AA vs AG + GG)	2.12 (-3.00–7.24)	0.413
	Recessive (AA + AG vs GG)	-13.76 (-40.75–13.23)	0.315
	Additive (GG vs GT vs TT)	0.36 (-5.73–6.44)	0.908
rs2934965	Dominant (GG vs GT + TT)	0.21 (-6.33–6.76)	0.949
	Recessive (GG + GT vs TT)	3.44 (-23.72–30.59)	0.803
	Additive (GG vs GA vs AA)	-1.75 (-5.65–2.15)	0.376
rs876493	Dominant (GG vs GA + AA)	-2.64 (-7.44–2.15)	0.278
	Recessive (GG + GA vs AA)	-0.06 (-9.45–9.34)	0.991
	Additive (CC vs CG vs GG)	-0.14 (-4.76–4.48)	0.952
rs2941523	Dominant (CC vs CG + GG)	0.95 (-4.26–6.16)	0.719
	Recessive (CC + CG vs GG)	-10.37 (-26.07–5.33)	0.194

^†Unstandardized regression coefficient.

Regression models are adjusted for covariates (age, sex and sickle cell type).

Table 4. . Association analyses of utilization scores with phenylethanolamine N-methyltransferase polymorphisms.

Single-nucleotide polymorphism	Model	Incidence risk ratio (97.5% CI)	p-value
	Additive (AA vs AG vs GG)	0.96 (0.66–1.41)	0.839
rs5638	Dominant (AA vs AG + GG)	0.98 (0.67–1.46)	0.930
	Recessive (AA + AG vs GG)	0.33 (0.01–5.97)	0.412
	Additive (GG vs GT vs TT)	0.61 (0.38–1.01)	0.044*
rs2934965	Dominant (GG vs GT + TT)	0.63 (0.38–1.06)	0.073
	Recessive (GG + GT vs TT)	0.15 (0.01–2.72)	0.164
	Additive (GG vs GA vs AA)	0.68 (0.51–0.93)	0.012*
rs876493	Dominant (GG vs GA + AA)	0.67 (0.47–0.97)	0.030*
	Recessive (GG + GA vs AA)	0.51 (0.24–1.15)	0.088
	Additive (CC vs CG vs GG)	0.68 (0.47–0.99)	0.033*
rs2941523	Dominant (CC vs CG + GG)	0.73 (0.49–1.09)	0.115
	Recessive (CC + CG vs GG)	0.12 (0.02–0.65)	0.017*

^†Regression models are adjusted for covariates (age, sex and sickle cell type).

*p-values < 0.05.

Strikingly, sex-specific analyses of utilization showed statistically significant associations in females with the rs876493 allele but not in males. Females had a 44% reduction in utilization when carrying either one or two risk alleles – GA or AA genotypes (p = 0.003, Table 5). Sex-specific analysis did not yield any significant associations for the other three SNPs in the dominant model, in agreement with our findings from the overall analysis.

Table 5. . Gender-based association analyses of utilization scores in the dominant model.

Single-nucleotide polymorphism	Sex	Total subjects	IRR (97.5% CI)	p-value
rs5638	M	45	1.1 (0.49–2.47)	0.819
	F	84	0.74 (0.48–1.15)	0.171
rs2934965	M	45	0.51 (0.19–1.51)	0.187
	F	86	0.58 (0.33–1.04)	0.060
rs876493	M	43	1.1 (0.53–2.28)	0.803
	F	82	0.56 (0.38–0.82)	0.003*
rs2941523	M	45	0.62 (0.24–1.76)	0.285
	F	86	0.76 (0.50–1.17)	0.203

Regression models are adjusted for covariates (age, sex, and sickle cell type).

*p-values < 0.05.

F: Female; IRR: Incidence risk ratio; M: Male.

In the assessment of linkage disequilibrium, we found rs2934965 and rs2941523 exhibit strong linkage, forming a haploblock (Figure 1). Since the two SNPs individually exhibited associations with utilization, the haplotype of the two SNPs was further analyzed with respect to CPI as well as utilization groups, including 1) subjects with utilization versus those without utilization, and 2) low (up to three events) versus high (four or more events) utilization groups. Upon haplotype analysis, no association was found to reach statistical significance for either CPI or the utilization categories (Table 6).

Figure 1. . Linkage disequilibrium plot of four phenylethanolamine N-methyltransferase single-nucleotide polymorphisms.

Haplotype block organization of four PNMT single-nucleotide polymorphisms generated from Haploview 4.2 using the standard color scheme with D′ values showing the linkage disequilibrium coefficient. Red = high D′ and high LOD, white = low D′ and low LOD; shades of pink/red = low D′ and high LOD.

LOD: Logarithm of the odds.

Table 6. . Haplotype analysis of haploblock rs2934965[G>T]- rs2941523[C>G].

Haplotype	Frequency (%)	0 versus 1 or more utilization		0–3 versus 4 or more utilization		CPI
		Odds ratio	p-value	Odds ratio	p-value	B†	p-value
GG	7	0.79	0.748	0.71	0.528	-0.73	0.827
TG	8	0.45	0.170	0.46	0.126	-0.30	0.922

^†Unstandardized regression coefficient.

Reference haplotype: GC (frequency = 85%).

Regression models are adjusted for covariates (age, sex and sickle cell type).

CPI: Composite pain index.

Using bioinformatics tool SNPinfo:FuncPred we found that rs2934965, rs876493 and rs2941523 are located at potential transcription factor binding sites (TFBS). We further explored the effect of nucleotide change on TFBS using PROMO (Figure 2). We found that for rs2934965 a change from G to T (variant minor allele) led to loss of putative binding sites retinoid X receptor α (RXR-α) and TEA domain family member 2 (ETF). For rs2941523 it was predicted that a change from C to G (variant minor allele) led to loss of putative binding site TFII–I (transcription factor II–I). The specific input sequence of rs876493, which was investigated, did not yield any change in TFBS upon nucleotide variation.

Figure 2. . Functional predictions for rs2934965 and rs2941523.

Changes in the transcription factor binding sites of rs2934965[G>T] and rs2941523[C>G] as predicted by PROMO. Nucleotide variations in the input sequence are indicated by red letter.

Discussion

In this study we studied the association of SNPs of PNMT with acute and chronic pain phenotypes in SCD. Production of epinephrine functions in part to maintain hemodynamic stability, though maladaptation and dysregulation secondary to polymorphisms are found in the literature [38–41].

Enhanced PNMT activity has been proposed to aggravate ischemic penumbras, thus precipitating painful episodes of VOC [6]. While epinephrine contributes to hyperalgesia [22], norepinephrine suppresses pain centrally [42]. Moreover, epinephrine can induce endothelial adhesion of sickle RBCs, consequently triggering VOC [23]. Since decreased transcription of PNMT would translate to reduced levels of epinephrine, it may be expected that polymorphisms reducing PNMT activity or expression would result in fewer or less severe incidences of VOC. As a consequence, such polymorphisms might favor lower acute pain or utilizations due to painful crisis. Correspondingly, our findings show that SCD subjects bearing minor alleles of rs876493, rs2934965 and rs2941523 in their genome experience acute pain crises less frequently. The lack of association of all four SNPs with chronic pain in SCD, despite their significant association with acute crisis pain, is not surprising given the different pathophysiology of the pain phenotypes. We have previously reported similar findings for the Val158Met polymorphism in COMT, another catecholamine-metabolizing enzyme [43]. Here it was found that while the Met allele showed significant association with acute crisis pain, chronic pain scores were not significantly different between the two alleles.

Of the three PNMT SNPs that showed significant associations with acute crisis pain, rs2934965 and rs2941523 lie within 2000 bases upstream of the 5′ end and rs876493 is located in the 5′UTR [16,44]. PNMT variants have also been studied in neurological disorders among non-Hispanic Caucasians. In Alzheimer’s disease (AD) the G allele of rs876493 significantly associated with early onset of AD [45]. In contrast to most reports of the A allele being protective, Mann et al. found that the AA genotype of the variant was in fact more prevalent in subjects with multiple sclerosis [46]. Our study found that the A variant of this SNP (rs876493) associated with lesser episodes of acute pain.

PNMT expression is highly regulated by transcription factors (TFs) and response elements in the promoter region [47,48]. Some of the TFs known to proximally regulate PNMT transcription include egr-1, Sp1, AP-2, glucocorticoid receptor and hypoxia-inducible factor-α [38,47–49]. Rodriguez-Flores et al. explored transcription factors binding at rs876493 SNP site and reported that the variant allele A of the SNP allowed for SOX17 binding and had an inhibitory effect on transcription [25]. Our finding that the A allele of rs876493 associated with decreased acute pain or utilization due to painful crisis is in agreement with this functional effect of the SNP.

It has been demonstrated that different haplotypes of SNPs in the gene’s 5′ flanking region and promoter region have significantly different transcriptional activity [16,25]. Our analyses of the SNP regions showed that not only does rs876493 lie in putative TFBS but rs2934965 and rs2941523 are also in or near a number of TFBS. In fact, our findings indicate that the minor allele of the rs2934965 (T allele) and rs2941523 (G allele) can cause potential loss of TFBS.

Interestingly, when we analyzed male and female subjects separately, the A allele of rs876493 conferred pain protection only in females. Sex is known to influence baseline PNMT activity by the cyclic and cascading effects of sex hormones [50,51]. But while this relationship is solidified among the literature, we find that studies have not been conducted for evaluating the role of sex in polymorphisms of the PNMT gene.

In our assessment of linkage disequilibrium for the four SNPs involved, we identified strong pair-wise linkage between rs2934965 and rs2941523. Analysis of this novel haploblock, however, did not reveal any significantly associated haplotypes. While most published studies evaluating PNMT polymorphisms have been performed in Caucasian or Asian populations, our study examined these SNPs in African-American subjects. One study looking at the role of PNMT SNPs in essential hypertension compared allelic frequencies between African-Americans and European-Americans, and found that the A allele of the variant rs876493 was more prevalent among whites [39]. Furthermore, they observed that rs876493 was in linkage disequilibrium with rs3764351 in both the African-American and European-American cohort. However, there was no significant association of the haplotypes with hypertension. On the contrary, this 2-SNP haplotype A-A was found to significantly associate with decreased risk of essential hypertension in a Han Chinese population [40]. A couple of other studies addressing Alzheimer’s disease and reward dependence have also reported the rs876493–rs3764351 linkage but failed to find significant associations with the haplotypes [45,52].

Conclusion

In summary, this is the first study reporting the association of PNMT polymorphisms with pain in SCD. We identified that allele A of SNP rs876493 in the 5′UTR of PNMT, previously implicated to alter transcriptional activity of the PNMT gene, significantly associated with reduced incidence of acute pain crises in SCD subjects. Similarly, 2kb upstream variants rs2934965 T allele and rs2941523 G allele also associated with decreased incidence of acute pain crises. Associations with chronic pain were, however, not significant. Bioinformatic analyses revealed that three (rs2934965, rs876493 and rs2941523) of the four SNPs were in or near putative TFBS. It was predicted that the variant minor alleles T of rs2934965 and G of rs2941523 may lead to loss of some of these TFBS. This warrants the need to perform further functional evaluation of these two understudied SNPs. Moreover, we found that while female patients carrying GA or AA genotype of rs876493 associated with significantly lower utilization scores compared with those who carried the GG genotype, the same was not true for males. We also found strong linkage disequilibrium between SNPs rs2934965 and rs2941523, though haplotypic associations with acute or chronic pain did not reach significance.

Additional studies with larger patient cohorts are needed to confirm these findings. Future studies should also account for the effect of medications such as hydroxyurea or glutamine or arginine supplementation on crisis frequency. While a number of subjects at our center participated in the hydroxyurea study, we currently do not have comprehensive medication prescription/use information for all of our study participants. Nonetheless, our findings indicate a possible role of genetic variants of PNMT on acute pain variability in SCD, thereby contributing to the growing evidence on the function of PNMT and epinephrine in the pathophysiology of this complex disease. Along with our previous findings in COMT, this study also indicates that polymorphisms in catecholamine-metabolizing enzymes may primarily influence acute pain in SCD.

Summary points.

• Phenylethanolamine N-methyltransferase polymorphisms rs876493 A allele, rs2934965 T allele and rs2941523 G allele significantly associated with decreased utilization.

• rs876493 A allele associated with significantly lower utilization only in female patients, not in males.

• rs2934965, rs876493 and rs2941523 lie in or near putative transcription factor binding sites.

• Variant alleles T of rs2934965 and G of rs2941523 may lead to loss of putative transcription factor binding sites.

• Associations with chronic pain phenotype of sickle cell disease were not significant.

• rs2934965 and rs2941523 were in strong linkage disequilibrium but haplotypic associations were not significant with either phenotype.

• This is the first evidence of association of phenylethanolamine N-methyltransferase polymorphisms with acute crisis pain in sickle cell disease.