Alexandria Journal of Medicine

ISSN: 2090-5068 (Print) 2090-5076 (Online) Journal homepage: https://www.tandfonline.com/loi/tajm20

Erythrocyte G6PD activity and GSH level as risk

factors for vascular complications among type 2
diabetics in Osogbo, Nigeria

Olaniyan O.O & Osadolor H.B

To cite this article: Olaniyan O.O & Osadolor H.B (2019) Erythrocyte G6PD activity and GSH
level as risk factors for vascular complications among type 2 diabetics in Osogbo, Nigeria,
Alexandria Journal of Medicine, 55:1, 95-99, DOI: 10.1080/20905068.2019.1688986

To link to this article: https://doi.org/10.1080/20905068.2019.1688986

© 2019 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group.

Published online: 12 Nov 2019.

Submit your article to this journal

Article views: 778

View related articles

View Crossmark data

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tajm20
ALEXANDRIA JOURNAL OF MEDICINE
2019, VOL. 55, NO. 1, 95–99
https://doi.org/10.1080/20905068.2019.1688986

Erythrocyte G6PD activity and GSH level as risk factors for vascular
complications among type 2 diabetics in Osogbo, Nigeria
Olaniyan O.Oa and Osadolor H.Bb
a
Department of Chemical Pathology, Osun State University, Osogbo, Nigeria; bDepartment of Medical Laboratory Science, University of
Benin, Benin City, Nigeria

ABSTRACT ARTICLE HISTORY

Background: Glucose-6-phosphate dehydrogenase (G6PD) catalyzes the oxidation of glucose- Received 26 July 2019
6-phosphate (G6P) to produce nicotine-adenosine-dinucleotide phosphate coenzyme Accepted 4 October 2019
(NADPH), a major defense against oxidative damage in erythrocytes. This study aims at KEYWORDS
evaluating enzymatic G6PD activity and erythrocyte glutathione (eGSH) as a biomarker of Glucose-6-phosphate
vascular complications among type 2 diabetics. dehydrogenase; glutathione;
Methods: Fasting Plasma glucose (FPG), glycated hemoglobin (HbA1c), erythrocyte G6PD type 2 diabetics; vascular
(eG6PD) activity and erythrocyte GSH (eGSH) level were determined along with measured complications
anthropometrics in 120 known type 2 diabetics (comprising 60 diabetics without vascular
complications and 60 diabetics with vascular complications) and 50 age and sex matched
apparently healthy non diabetes individuals recruited for this study.
Results: Result revealed significant reduced eG6PD activity and eGSH levels in type 2 diabetics
especially among those with vascular complications as compared to subjects without diabetes,
with a slight correlation between eGSH level and eG6PD activity of diabetics with vascular
complications.
Conclusion: The study, therefore, advocates measurement of G6PD activity and GSH level in
type 2 diabetics to properly monitor the progress of the disease.

1. Introduction 6-phosphate (G6P) through the hexokinase/glucoki-

nase enzymes. The formed G6P may be used either in
Glucose-6-phosphate dehydrogenase (G6PD EC
glycolysis to produce energy in the form of adenosine
1.1.1.49) enzyme is a rate-limiting/housekeeping cyto-
triphosphate (ATP) and NADH, used to store energy
plasmic enzyme of the pentose phosphate pathway
in the form of glycogen, or used by the PPP (otherwise
(PPP) that catalyzes the oxidation of glucose-6-phos-
known as the hexose monophosphate shunt) to form
phate to 6-phosphogluconolactone and concomitant
ribose-5-phosphate, which is required for nucleic acid
reduction of NADP+ to nicotinamide adenine dinu-
(ribose) synthesis, and nicotinamide adenine dinu-
cleotide (NADPH), which represents the only source
cleotide phosphate (NADPH) generated from NADP
of NADPH in erythrocytes [1]. G6PD is widely dis-
by G6PD enzyme. NADPH is the main cellular anti-
tributed in many species ranging from prokaryotes to
oxidant (reductant) defense mechanism found in the
eukaryotes with sequence identity resembling each
body [2].
other, in humans the G6PD encoding gene is located
NADPH, a hydrogen carrier is required by many
at the telomeric region of the long arm of the
essential cellular systems like glutathione recycling,
X chromosome (band Xq28) with 13 exons, 12
nitric oxide synthesis, cytochrome p450 system, and
introns, and length of 18kb [2,3]. The polypeptide
others for proper functioning and there is increas-
sequence of the dimer (sometimes tetramer) enzyme
ing evidence that G6PD activity is of major impor-
with about 515 amino acids contains conserved region
tance for NADPH production for defense against
of an octapeptide lysine residue (for enzymatic activ-
oxidative stress rather than for ribose production.
ity), heptapeptide dinucleotide (for binding activity)
Diabetes mellitus, the most common endocrine dis-
and a pentapeptide near the substrate binding site.
ease of carbohydrate metabolism is being consid-
Mutation of the G6PD gene results in protein variants
ered a free radical disease owing to the increase in
with different levels of enzyme activity and a wide
the production of free radical [6]. Diabetes mellitus
spectrum of biochemical and clinical phenotypes [4,5].
causes an increase in metabolic flux of the polyol
Traditionally, digested glucose from carbohydrate
pathway with increase cellular demand for NADPH
metabolism is transported into cells from plasma to
by the NADPH-dependent aldose reductase and
undergo phosphorylation forming glucose-
subsequent deficiency generation of the endogenous

CONTACT Olaniyan O.O [email protected] Department of Chemical Pathology, College of Health Sciences, Osun State University, Osogbo,
Osun State, Nigeria
© 2019 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits
unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
96 OLANIYAN O.O. AND OSADOLOR H.B

non-enzymatic antioxidant (glutathione) defense Nephropathy

mechanisms. Therefore, this study aims at evaluat- Retinopathy
Cataract
ing enzymatic G6PD activity and GSH level as risk 1.67%
factors for the development of vascular complica-
tions among type 2 diabetics. 25.83%
50.00%

2. Materials and methods

2.1. Subjects 11.67%

This case-control hospital-based study was conducted 6.67%

at the State Specialist Hospital, Osogbo, Southwest
1.67% 3%
Nigeria, using a simple non-probability sampling
technique, the study protocol was in accordance with Figure 1. Patient distribution in the diabetes group.
Helsinki declaration and approved by the Ethics
Committee of the Osun State University. One hundred
and twenty type 2 diabetics between 30 and 65 years
old were recruited from the Hospital diabetes clinic. branch block while, diabetic foot (two subjects) is
Type 2 diabetics were those previously diagnosed in established by any pathological features of the foot
the Hospital using WHO standard [7], and classified resulting directly from diabetes mellitus. The remain-
according to the presence or absence of diabetes vas- ing 60 diabetics were those without any vascular com-
cular complications. Sixty of the type diabetics were plications (DM-VC) (Figure 1). Exclusion criteria
with vascular complications (DM+VC), vascular com- include type 1 diabetics or secondary diabetics, type
plications were defined as any documented vascular 2 diabetics using oral hypoglycemics other than bigua-
disease (also diagnosed previously in the hospital) at nides, smokers, recent surgery and acute or chronic
least a year after the diagnosis of type 2 diabetes using infection or undergoing isotope diagnostic or irradia-
standard medical protocols. Diabetes eye complica- tion therapy. The control subjects were 50 apparently
tions (retinopathy-31 subjects, cataract-8 subjects) healthy subjects recruited among the staff of Osun
were diagnosed after dilated pupil ophthalmoscopic State University but with similar age and sex distribu-
examination in a darkened room [8], diabetes periph- tion to those of the diabetes patients. They were sub-
eral neuropathy (14 subjects) was diagnosed based on jected to complete medical examination to exclude the
the presence of two or more signs of paresthesia, presence of any medical problems. Informed consent
absent pinprick, light touch sensation, sense of posi- was obtained from all the participants before the com-
tion and absent tendon reflexes or muscular atrophy. mencement of the study while, a structured research
Overt nephropathy (two subjects) was based on the questionnaire was administered on each subject to
criteria that a patient repeatedly had either a urinary collect relevant demographic diabetes information.
albumin excretion rate of >200 µg/min or a positive In addition, anthropometric data to calculate body
urinalysis for protein using a reagent strip [9,10]. Type mass index (BMI) and blood pressure were collected
2 diabetes ischemic heart disease (three subjects) was using standardized methods. The clinical characteris-
indicated based on the standard 12 lead electrocardio- tics of the diabetics and control subjects are as indi-
gram with the presence of Q wave or of left bundle cated in Table 1.

Table 1. Demographic and laboratory data of the study population.

Parameters Controls Type 2 diabetics DM˗VC DM+VC
Sex (♂/♀) 20/30 45/75 25/35 20/40
Age (years) 54.16 ± 9.61 53.38 ± 9.38 53.97 ± 9.71 53.35 ± 8.41
DD (years) NA 4.55 ± 2.91 4.43 ± 2.55 6.67 ± 2.84
Family History with DM 5 [10] 67 (55.83) 35 (58.33) 32 (53.33)
BMI (kg/m2) 20.91 ± 0.85 25.71 ± 4.13 24.86 ± 3.43 26.56 ± 4.61
Obese (%) NA 13 (10.83) 5 (8.33) 8 (13.33)
Overweight (%) NA 47 (39.17) 20 (33.33) 27 (45)
Normal weight (%) 50 (100) 60 (50) 35 (58.34) 25 (41.67)
SBP (mmHg) 117.00 ± 13.48 130.62 ± 8.51 130.65 ± 8.71 130.58 ± 8.38
DBP (mmHg) 83.70 ± 4.77 88.93 ± 5.22 87.73 ± 5.16 90.13 ± 5.05
Hypertensive (%) NA 13(10.83) 4 (6.67) 9 [15]
G6PD deficiency (%) 3 [6] 7 (5.83) 4 (6.67)3 [5]
Result expressed as mean, standard deviation and number (percentage)
DD = Diabetes duration
BMI = Body mass index
SBP = Systolic blood pressure
DBP = Diastolic blood pressure
 ALEXANDRIA JOURNAL OF MEDICINE 97

2.2. Blood samples and laboratory analysis vascular complications (4.43 ± 2.55). DM+VC group had
a significantly higher frequency of systemic hypertension
Fasting blood sample was obtained by venipuncture into
and significantly higher levels of systolic and diastolic
fluoride oxalate (2 mL) and heparinized (3 mL) tubes for
blood pressure and plasma glucose than the control sub-
the biochemical estimation of fasting plasma glucose-
jects. The Mean FPG and HbA1c in type 2 diabetics were
FPG estimated using enzymatic glucose oxidase/hydro-
6.92 ± 1.74mM and 7.43 ± 1.49%, respectively, as com-
lysis method, glycosylated hemoglobin-HbA1c using ion
pared to control subjects (4.69 ± 0.71mM and 4.96 ±
exchange resin method. Methemoglobin reduction test as
0.69%) were significantly (p < 0.05) high, indicating
described by Amiwero and Olatunji [11] was used to
poor glycemic control among diabetics (Table 1). All
screen all subjects for G6PD deficiency. Erythrocyte
the G6PD-deficient subjects had undetectable levels of
Glucose-6-Phosphate dehydrogenase (G6PD) activity
enzymatic red blood cell activity (total deficiency) and
was determined based on the enzyme ability to reduces
none of the subjects (either control or diabetics) with
NADP to NADPH, and the rate of reduction of NADP+
G6PD deficiency had previously been treated for G6PD
was measured at 340nm while erythrocyte glutathione
enzyme deficiency in the past 2 years before their enroll-
was determined using modified method of Chakrabarty
ment in this study.
et al. [12] where glutathione in the protein-free red cell
The DM-VC and DM+VC had higher FPG and
lysates was made to react with DTNB solution and the
HbA1c levels of 5.38 ± 0.38mM, 6.16 ± 0.53% and
complex formed read at 412nm. Plasma total antioxidant
8.46 ± 0.45mM, 8.69 ± 0.68%, respectively, than the
capacity (pTAC), was also measured on the basis of the
control group (4.69 ± 0.71mM, 4.96 ± 0.69%).
ability of plasma antioxidants plasma to reduce Fe3+-
However, no significant difference (p > 0.05) was
TPTZ to Fe2+-TPTZ [13].
recorded in the FPG of DM-VC group compared to
the control individuals, but a significant (p < 0.05)
2.3. Statistical analysis twofold rise in the FPG level of the DM+VC group
was recorded compared to the normal individuals.
Normal distribution was analyzed with Kolmogorov Therefore, the DM+VC and DM-VC were considered
Smirnov test while Kruskal–Wallis test was used if to be under poor (PDC) and good (GDC) glycemic
parameters did not follow normal distribution. diabetes control, respectively, at the time of the experi-
Differences between the groups were tested with ment. Conversely, the level of HbA1c was significantly
t-test for independent variables or Mann-Whitney-U (p < 0.05) increased in DM-VC and DM+VC diabetic
test for nonparametric variables. Pairwise compari- groups (Table 2).
sons between multiple groups were analyzed with one- The eG6PD enzymatic activity, eGSH and pTAC
way analysis of variance (ANOVA). Correlations were levels of type 2 diabetics (6.18 ± 1.85mU/gHb, 47.64 ±
analyzed with Pearson’s correlation coefficient or 18.34mg/gHb and 0.76 ± 0.24mM) respectively, were
Spearman correlation for nonparametric variables. significantly (p < 0.05) lowered than control (8.28 ±
Significance was assumed at p < 0.05. All statistical 0.57mU/gHb, 66.23 ± 3.55mg/gHb and 1.48 ± 0.33mM)
analyzes were performed using Graph-Pad prism ver- group. eG6PD activity, eGSH and pTAC levels of DM
sion 8 software. +VC (PGC) group (4.52 ± 0.88 mU/gHb, 29.66 ±
3.42mg/gHb and 0.64 ± 0.22mM) were significantly (p
< 0.05) decreased two-folds compared to the control
3. Result group. The DM-VC (GGC) group also has a statistically
The study group consisted of 120 patients (mean age: (p < 0.05) similar but reduced eG6PD activity, and pTAC
53.38 ± 9.38 years), DM-VC group were 60 individuals level (7.88 ± 0.69mU/gHb, 0.89 ± 0.18mM) to the control
(mean age: 53.97 ± 9.71 years) and DM+VC group were group, while eGSH level (65.61 ± 3.15 mg/gHb) of the
also 60 subjects (mean age: 53.35 ± 8.41 years). The mean DM-VC (GGC) did not differ statistically (p > 0.05) to the
diabetes duration was 4.55 ± 2.91years in all patients as control group (Table 2). Correlation analyses revealed no
diabetics with vascular complications has the highest association between glycemic control parameters and
mean diabetes duration (6.67 ± 2.84) than those without G6PD activity in the diabetes group, but a slight

Table 2. Biochemical levels of glycemic control, G6PD activity, GSH and TAC levels of the study population.
FPG HbA1c G6PD activity GSH TAC
Group (mmol/L) (%) (mU/gHb) (mg/gHb) (mM)
Control subjects 4.69 ± 0.71a 4.96 ± 0.69 a 8.39 ± 0.61a 66.23 ± 3.55a 1.48 ± 0.33a
Type 2 Diabetics 6.92 ± 1.74b 7.43 ± 1.49 b 6.18 ± 1.85b 47.64 ± 18.34b 0.76 ± 0.2 b
DM -VC 5.37 ± 0.59a 6.16 ± 0.53 c 7.88 ± 0.60c 65.61 ± 3.15a 0.89 ± 0.18c
DM +VC 8.46 ± 0.45c 8.69 ± 0.96 d 4.52 ± 0.88d 29.66 ± 3.24c 0.64 ± 0.22d
Result expressed as mean ± standard deviation
Values with different subscripts (a, b, c, d) down a column are significantly different from each other (P < 0.05).
DM-VC and DM+VC denote diabetic group without vascular and with vascular complications, respectively.
98 OLANIYAN O.O. AND OSADOLOR H.B

r = 0.2659 a decrease level of reduced glutathione which may

75 become critical upon oxidative stress moreover, evi-
dence from both in-vivo and in-vitro studies had
70 shown that glutathione depletion enhances NADPH-
dependent lipid peroxidation in diabetics.
GSH

65

60 5. Conclusion

55
The novelty of this study was the evaluation of G6PD
6 7 8 9 10 activity and glutathione level in diabetics with and with-
G6PD out vascular complications. Uncontrolled hyperglycemia
in diabetes leads to vascular complications with conco-
Figure 2. G6PD activity as a function of GSH level in plasma of mitant decrease in G6PD enzyme activity and reduced
DM-VC group.
glutathione level. Evaluation of G6PD activity and glu-
tathione level can, therefore, predict diabetes injury due
correlation (r = 0.2659) between G6PD activity and GSH to inappropriate oxidation/anti-oxidation process in type
level was found among the DM-VC (GGC) group 2 diabetics. We, therefore, advocate large-scale longitudi-
(Figure 2). nal study to validate the result of this study.

4. Discussion Disclosure statement

In the present investigation, the overall G6PD deficiency No potential conflict of interest was reported by the authors.
prevalence from our study was 5.88%, a value lower to
previously reported deficiency among adults in our local-
ity [14]. Diabetic patients from our study have 5.83% Notes on contributors
deficiency prevalence [of these, three (5.0%) were from
diabetics with vascular complications and four (6.67%) Olayinka Olaolu Olaniyan is a Medical Laboratory Scientist
at Osun State University, Osogbo, and currently undergoing
were from diabetics without vascular complications] doctoral training at the University of Benin, Benin City,
while control subjects have 6% deficiency prevalence, Nigeria.
a lower prevalence value as compared to Engwa et al.
Humphrey Benido Osadolor is a Professor of Clinical
[15], who did similar prevalence study of G6PD defi- Chemistry at the Department of Medical Laboratory
ciency among type 2 diabetics. Science, University of Benin, Benin City, Nigeria. His area
This study also revealed that the mean enzymatic of expertise is in the field of environmental toxicology and
activity of G6PD is significantly lower in diabetics human health
than nondiabetes subjects and also significantly lowered
in diabetics with vascular complications than those
without vascular complications. The reason for the References
decreased activity of the enzyme might be due to the [1] Mahmoud AA, Nor El-Din AKA. Glucose-6-phos-
reported hyperglycemia which causes activation of pro- phate dehydrogenase activity and protein oxidative
tein kinase-A and subsequent phosphorylation of modification in patients with type 2 diabetes
G6PD enzyme thereby lowering the enzyme activity mellitus. J Biomarkers. 2013;2013:1–7.
[1,6,16–19]. [2] Stanton RC. Glucose-6-phosphate dehydrogenase,
NADPH, and cell survival. IUBMB Life; 2012;64(5):
Likewise, our study revealed a reduced red cell glu- 362–369.
tathione level in type 2 diabetics as compared to control [3] Rai V, Kumar P. Epidemiological study of
subjects. The decreased level is more pronounced glucose-6-phosphate dehydrogenase deficiency in
among type 2 diabetics with vascular complications scheduled caste population of India. J Anthropol.
than those without vascular complications, supporting 2012;2012:1–5.
[4] Kotaka M, Gover S, Vandeputte-Rutten L, et al.
the hypothesis of Likidlilid et al. [20], who stated that
Structural studies of glucose-6-phosphate and
the decreased red cell glutathione may be due to high NADP+ binding to human glucose-6-phosphate
rate of oxidative stress in diabetics caused by reduced dehydrogenase. Acta Crystallogr D Biol Crystallogr
pentose phosphate pathway activity or increased glu- [Internet]. 2005 May;61(Pt 5):495–504. [cited 2019 Jul
tathione peroxidase activity as glutathione system pre- 22]. Available from: http://www.ncbi.nlm.nih.gov/
vents oxidative stress derangement in erythrocytes. This pubmed/15858258
[5] Gómez-Manzo S, Terrón-Hernández J, De la Mora-
study however, in agreement with Bozzi et al [21]., who De la Mora I, et al. The stability of G6PD is affected by
showed a link between low G6PD enzymatic activities mutations with different clinical phenotypes.
leading to an inefficient NADPH generating system and Int J Mol Sci. 2014;15:21179–21201.
 ALEXANDRIA JOURNAL OF MEDICINE 99

[6] Rashidi H, Shafiei M, Hamidian R. Erythrocytic glu- blood donors and jaundiced neonates in Osogbo,
cose-6-Phosphate. Pak J Med Sci. 2009; 25(4): Nigeria. J Med Lab Diagnosis. 2010;1(1): 1–4..
665–668. [15] Engwa GA, Nwalo FN, Chibuzor GE, et al. Relationship
[7] Definition and diagnosis of diabetes mellitus and between type 2 diabetes and glucose-6-phosphate dehy-
intermediate hyperglycemia [Internet]. World health drogenase (G6PD) deficiency and their effect on oxida-
organisation; 2006. [cited 2018 Jun 20]. Available tive stress. J Diabetes Metab. 2018;09(08): 1–8.
from: https://www.who.int/diabetes/publications/ [16] Zang Z, Apse K, Pang J, et al. High glucose inhibits
Definition and diagnosis of diabetes_new.pdf glucose-6-phosphate dehydrogenase via cAMP in
[8] Frank RN. Diabetic retinopathy. New Eng. J. Med. aortic endothelial cells. J Biol Chem.
2004;350(01): 48–58. 2000;275:40042–40047.
[9] Nakhoul FM, Zoabi R, Kanter Y, et al. Haptoglobin [17] Xu Y, Osborne BW, Stanton RC. Diabetes causes inhibi-
phenotype and diabetic nephropathy. Diabetologia tion of glucose-6-phosphate dehydrogenase via activation
[Internet]. 2001 May;44(5):602–604. [cited 2019 Jul 15]. of PKA, which contributes to oxidative stress in rat kidney
Available from: http://www.ncbi.nlm.nih.gov/pubmed/ cortex. Am J Physiol Physiol. 2005;289:F1040-F1047.
11380078 [18] Zhang Z, Liew CW, Handy DE, et al. High glucose
[10] Gross JL, De Azevedo MJ, Silveiro SP, et al. Diabetic inhibits glucose-6-phosphate dehydrogenase, leading
Nephropathy : Diagnosis, Prevention, and to increased oxidative stress and β-cell apoptosis.
Treatment’, Diabetes Care, 2005;28(1): 176–188. Faseb J. 2010;24:1497–1505.
[11] Amiwero CE, Olatunji PO. Re-evaluation of methae- [19] Wan GH, Tsai SC, Chiu DTY. Decreased blood activ-
moglobin reduction as a screening procedure for ity of glucose-6-phosphate dehydrogenase associates
glucose-6-phosphate dehydrogenase (G6PD). Afr with increased risk for diabetes mellitus. Endocrine.
J Med Med Sci. 2007;36(2): 177–181. 2002;19:191–196.
[12] Chakrabarty J, Vidyasagar MS, Fernandes D, et al. [20] Likidlilid A, Patchanans N, Poldee S, et al.
Effectiveness of pranayama on the levels of serum Glutathione and glutathione peroxidase in type 1
protein thiols and glutathione in breast cancer diabetic patients. J Med Assoc Thail. 2007;90(9):
patients undergoing radiation therapy: a randomized 1759–1767 [cited 2019 Jul22]: Available from
controlled trial. Indian J Physiol Pharmacol. 2013;57 http://www.ncbi.nlm.nih.gov/pubmed/17957916.
(3):225–232. [21] Bozzi A, Parisi M, Strom R. Erythrocyte glutathione
[13] Benzie IFF, Strain JJ. The Ferric Reducing Ability of determination in the diagnosis of
Plasma (FRAP) as a measure of “‘Antioxidant Power’”: glucose-6-phosphate dehydrogenase deficiency.
the FRAP Assay. Anal Biochem. 1996;239:70–76. Biochem Mol Biol Int [Internet]. 1996 Oct;40
[14] Akanni EO, Oseni A, Agbona VO, et al. (3):561–569 [cited 2019 Jul 22]: Available from
Glucose6phosphate dehydrogenase deficiency in http://www.ncbi.nlm.nih.gov/pubmed/8908366