Oxidative stress is a disturbance in the pro-oxidant [reactive oxygen species (ROS)] and antioxidant balance in favour of the pro-oxidant state [1, 2]. Increased oxidative stress has been found in people with diabetes, presumably contributing to diabetic micro- and macroangiopathic complications [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]. One form of oxidative damage to cellular macromolecules, mutations in mitochondrial DNA (mDNA) have been noted as a possible consequence of DNA damage [13, 15, 16]. Most experiments on diabetes mellitus have been directed to large mDNA deletion [13, 15], and experiments on somatic point mutations have been very limited [16]. Somatic point mutations have also been investigated in connection with aging or age-related diseases [17, 18, 19, 20].

MDNA has a higher frequency of somatic mutations than nuclear DNA (nDNA). Deamination and methylation are thought to bring about transition mutations, i.e. the substitution of one purine or one pyrimidine base for its respective other, in nDNA and mDNA [21, 22, 23]. Both processes often occur in vivo [24]. Transversion mutations occur in the same way through oxidative damage caused by 8-hydroxy-2′-deoxyguanosine (8-OHdG) [25, 26, 27, 28, 29, 30]. Along with these mutations, mitochondria are also likely to have fewer repair enzymes, e.g. glycosylase, for base excision repair than the nucleus has [31], and no nucleotide excision repair capacities [32]. MDNA is also located near the respiratory chain, potentially generating ROS. In addition, mDNA has no histone proteins, which protect DNA from deamination, methylation or oxidation [23, 33]. As a result, mDNA is thought to have a higher mutability and subsequently acquire 5 to 20 times as many somatic mutations as nDNA [34, 35].

Because of its abundance and mutagenicity 8-OHdG has been reported to be the most important indicator of oxidative DNA damage [5, 25, 36]. This 8-OHdG comes from an oxidative reaction to the guanine (G) base, which is highly susceptible to oxidative stress, since it has the lowest oxidation potential among bases. The substance 8-OHdG promotes the formation of Watson-Crick pairs not only with cytosine (C), but also with adenine (A), and can give rise to G:C site to thymine (T):A site or T:A site to G:C site transversion point mutations (substitution of a purine base for a pyrimidine base or vice-versa; referred to below as “8-OHdG-associated mutations”) [25, 37, 38]. The amount of 8-OHdG is reportedly larger in the DNA and urine of people with diabetes than in those of non-diabetic people [5, 7, 9]. Urinary 8-OHdG is a familiar marker of oxidative stress, because it is quite stable in the urine and not easily affected by foods [39, 40].

We therefore attempted to identify the kinds of mutations that can actually be induced in the mDNA of patients with diabetes mellitus, while concurrently assaying urinary 8-OHdG.

Subjects and methods

Subjects

Patients were recruited from the outpatient department of the Shinshu University Hospital. For the control group non-diabetic subjects of the same sex and age were selected at random. All participants gave their written informed consent before enrolling in the study, which was approved by the Ethics Committee of Shinshu University School of Medicine, Matsumoto, Japan. The type of diabetes was diagnosed according to the 1998 criteria of the World Health Organization [41]. Diabetic retinopathy diagnosed by funduscopy was designated as a marker of microangiopathy. One patient had clinically overt macroangiopathy (cerebral infarction). There were no active smokers among our subjects, whose characteristics are shown in Table 1. Postprandial peripheral blood and urine samples were taken from each subject. The blood samples were subjected to DNA extraction and urine samples were frozen at −20°C until analysis.

Table 1 Characteristics of diabetic and non-diabetic control subjects
Full size table

DNA extraction and subcloning

DNA was extracted from peripheral blood cells using a DNA Extractor WB kit (NaI method; Wako Pure Chemical Industries, Osaka, Japan). The extracted DNA (0.5 to 1.0 µg) was subjected to a PCR with Taq DNA polymerase (Takara Bio Shiga, Japan) to amplify two segments of mDNA (GenBank J01415), nt 15997 to 16400 (forward primer, 5′-CCA CCA TTA GCA CCC AAA GC-3′: backward primer, 5′-TGA TTT CAC GGA GGA TGG TG-3′), and nt 3150 to 3403 (forward primer, 5′-AGG ACA AGA GAA ATA AGG CC-3′: backward primer, 5′-CAC GTT GGG GCC TTT GCG TA-3′). The former (segment A) is a hypervariable segment, the latter (segment B) is a segment encompassing the t-RNA Leu (UUR) gene. PCR was done in only 12 cycles consisting of denaturation at 94°C, annealing at 55°C and extension at 72°C for 1 min each. The PCR products were electrophoresed on a 1.5% (wt/vol) agarose gel with a Ready-to-Load 100 bp ladder (Invitrogen, Carlsbad, Calif., USA) running in parallel for approximate PCR product band sizing and stained by ethidium bromide for 15 min. Proper bands at the size of 444 bps (segment A) or 294 bps (segment B) were then cut out, and the PCR products were purified by a GFX PCR DNA and Gel Band Purification Kit (Amersham Pharmacia Biotech, Piscataway, N.J., USA). The purified PCR products were ligated into a pCR2.1 vector and transformed into competent cells of Escherichia coli by using an Original TA cloning Kit (Invitrogen). The transformed cells were then spread onto LB agar plates containing 143.1 µmol/l of ampicillin with 40 µl of 5-Bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-Gal) stock solution (48.9 mmol/l).

Sequencing of mDNA

Recombinant plasmids isolated from 20 white colonies per patient or control subject were subjected to sequencing with a dye terminator cycle sequencing kit (Perkin-Elmer, Foster, Calif., USA). The array of mDNA sequences was determined by using the ABI PRISM 310 Genetic Analyzer (Applied Biosystems, Foster, Calif., USA) and comparing carefully with the revised Cambridge Reference Sequence [42]. The mutations identified in all clones (homoplasmy) are referred to as “germ-line mutations” and the other heteroplasmic mutations as “somatic mutations”. The two A3243G transition mutations from 20 colonies in patient DM_3 (Table 1) were the only exceptions. They were qualified as germ-line mutations in heteroplasmy, because her daughter had been found to have mitochondrial diabetes.

Urinary 8-OHdG assay

The 8-OHdG in the urine was measured by enzyme-linked immunosorbent assay [43]. The intra- and inter-assay coefficients of variation were 6.0% and 8.6% respectively. We calculated the urinary 8-OHdG by dividing 8-OHdG by creatinine per unit volume of urine.

HL-60 cell culture

The HL-60 human promyelocytic/myeloblastic cell line was purchased from the American Type Culture Collection (Rockville, Md., USA) and maintained in a RPMI 1640 medium (11.1 mmol/l glucose; Nikken BioMedical Laboratory, Kyoto, Japan), supplemented with 10% fetal bovine serum (Life Technologies, Grand Island, N.Y., USA) and 1% antibiotic-antimycotic (100 U/ml penicillin G sodium, 171.9 µmol/l streptomycin and 270.5 nmol/l amphotericin B; Invitrogen). The HL60 cells were divided into two groups, which were further cultured with glucose concentrations of 5.6 and 27.8 mmol/l respectively, by adding phosphated-buffered saline and/or a concentrated glucose solution. After 6 weeks of incubation, somatic point mutations in mDNA were assessed as described earlier.

Mathematical modelling of PCR error

Mathematical modelling of the PCR error [18] was as follows: Let f be the PCR polymerase error rate per base, L the length of the strand to be amplified, S0 the initial number of strands and η the fraction of molecules replicated in each cycle. The probability of introducing k errors while copying a strand is:

$$p(k) = C(L,k)f^{k} (1 - f)^{{L - k}} .$$

Let S(c,k) be the number of strands present after c cycles with k PCR-induced errors. The fraction of strands present after c cycles with k errors is:

$$F(c,k) = S(c,k)/{\left( {S_{0} (1 + \eta )^{c} } \right)} = {\left[ {1/(1 + \eta )} \right]}{\left[ {F(c - 1,k) + \eta {\sum\nolimits_{j \leqslant k} {F(c - 1,j)p(k - j)} }} \right]}.$$

Therefore, the average number of errors per strand is:

$$ < {\text{k}} > = {\sum\nolimits_{\text{k}} {F(c,k)k} } = cfL\eta /(1 + \eta ).$$

The overall PCR error per base is:

$$ < {\text{k}} > /L = cf\eta /(1 + \eta ).$$

Data analysis

The mutational burden for a given sample was calculated as the number of mutations found divided by the total number of bases examined. The data were reported as the mean ± SEM. The correlation between variables was calculated by the linear regression analysis of untransformed values. A p value lower than 0.05 was regarded as representing a statistically significant difference. The Mann-Whitney U-test was used to compare quantitive data from two segments of mDNA and two groups. Fisher’s exact test was used to compare the ratios of the different kinds of somatic transition mutations. All analyses were done using Statview 5.0J (Abacus Concepts, Berkeley, Calif., USA).

Results

The PCR error per base calculated during 12 PCR cycles was ~6.00×10−4/bp, with an error rate of ~10−4 per base. To test the overall PCR error per base, we experimentally measured PCR errors by using a single bacterial colony as a template for PCR. Sequencing of 25 recombinant plasmids for segment A and 32 for segment B revealed no point mutations, indicating that the overall PCR error per base in the 12 PCR cycles was less than 5.49×10−5/bp. When all the molecules had been replicated during each cycle, the PCR error rate was less than 9.15×10−6/bp.

The frequency of germ-line mutations (polymorphisms) was significantly higher in segment A than in segment B (49.51±5.90×10−5 mutations per bp for A, 1.79±1.79×10−5 mutations per bp for B; p<0.0001, according to Mann-Whitney U-test). However, no differences were found between segments A and B in the frequency of (i) total somatic mutations (44.35±9.57×10−5 mutations per bp and 59.06±10.94×10−5 mutations per bp; Table 2), (ii) transversion mutations (6.19±3.23×10−5 and 10.74±5.55×10−5), (iii) transition mutations (37.13±10.22×10−5 and 44.74±9.23×10−5), and (iv) insertions and deletions (1.03±1.03×10−5 and 3.58±2.40×10−5).

Table 2 All somatic mDNA mutations and urinary 8-OHdG
Full size table

Since there were no differences in the mutational burden between the two segments, we assessed the aggregate somatic mutational burden by combining segments A and B. None of the somatic mutations detected in the twenty clones were the same, suggesting that the degree of heteroplasmy was less than 5%. There were no significant differences in the total somatic point mutational burden between the diabetic and control group (49.39±8.50×10−5 mutations per bp in patients and 51.93±12.00×10−5 mutations per bp in healthy control subjects; p=0.58).

However, the aggregate somatic transversion mutational burden in diabetic patients was significantly higher than in the control subjects (13.93±4.57×10−5 mutations per bp vs 1.27±1.27×10−5 mutations per bp; p=0.031; Fig. 1a), while no significant differences were found when two segments were analysed separately (p=0.30 and 0.068, in segments A and B respectively). The frequency of somatic transition point mutations was not different between the two groups (31.66±6.63×10−5 vs 50.65±12.35×10−5; p=0.17; Fig. 1b). Interestingly, insertions (1) and deletions (2) were detected in only three diabetic patients. In addition, the linear regression analysis showed a significant correlation between aggregate somatic transversion mutational burden and HbA1c values in diabetic patients (r=0.840, r 2=0.706, p=0.036; Fig. 2). The HbA1c values were 8.55±1.03% in patients with diabetes mellitus and 5.13±0.12% in the control subjects (p<0.005). Moreover, two transversion mutations including one “8-OHdG-associated mutation” were detected from 30 clones of mDNA from HL60 cells cultured at a glucose concentration of 27.8 mmol/l. At a glucose concentration of 5.6 mmol/l one transition mutation was found (Fig. 1a, b).

Fig. 1a, b
figure 1

Aggregate incidence of somatic mutations in leucocyte mDNA from control subjects (c1–c6), diabetic patients (D1–D6) and HL-60 grown for 3 weeks at glucose concentrations of 5.6 and 27.8 mmol/l. The aggregate transversion mutational burden (number of transversion mutations found in multiple clones, divided by the total number of bps sequenced per subject) is shown (a) for each of six samples from control subjects and from patients with diabetes, and for one sample of HL-60 per glucose concentration. The black bar represents 8-OHdG-associated transversion mutation. The double-hatched bar represents the other transversion mutations between Watson-Crick base pair; like A:T site to T:A site, or G:C site to C:G site, or vice-versa. The aggregate transition mutational burden (number of transition mutations found in multiple clones, divided by the total number of bps sequenced per subject) is shown in a similar way (b)

Full size image
Fig. 2
figure 2

HbA1c values and somatic transversion mutational burden of mDNA in peripheral blood cells from control subjects and diabetic patients. The white circles represent total transversion in control subjects. The black circles represent total transversion in diabetic subjects. Among diabetic patients there is a positive correlation between the incidence of somatic transversion mutations and HbA1c values (r=0.840, r 2=0.706, p=0.036; Y=−6.428*10−4+9.147*10−5X, according to linear regression analysis)

Full size image

The amount of 8-OHdG in the urine tended to be slightly larger in diabetic patients than in control subjects (18.46±3.75 ng/mg creatinine compared with 12.43±1.39 ng/mg creatinine). Urinary 8-OHdG was not correlated with the amount of transversion or transition point mutations of mDNA in any of the subjects (Fig. 3). Also no relation was found between age and HbA1c values. We only observed 8-OHdG-associated transversion mutations in three patients with diabetes mellitus, and at 22.64±6.54 ng/mg creatinine their average concentration of urinary 8-OHdG was slightly higher than in the other diabetic patients.

Fig. 3
figure 3

Urinary 8-OHdG and somatic transversion mutational burden in leucocyte mDNA from control subjects and diabetic patients. Including all samples from all subjects, there is a negative correlation between the total number of transversion mutations, 8-OHdG-associated transversion mutations and urinary 8-OHdG. The white square represents total transversion in control subjects. The black square represents total transversion in diabetic subjects. 8-OHdG-associated transversion mutations in control subjects are depicted by white circles. In diabetic patients they are depicted by black circles

Full size image

Discussion

By means of a method similar to one already used [18], we directly assessed the aggregate burden of somatic point mutations in two segments of mDNA by using a PCR-cloning-sequencing strategy to detect acquired somatic point mutations. In our study it was necessary to take the PCR error into account. Taq DNA polymerase has an assumed error rate of ~10−4 per base per doubling, resulting mainly in A:T to G:C transition mutations. Other artefacts such as heat-induced mutagenesis can also be created during PCR [44]. To keep the PCR error per base at a minimum, we used only 12 PCR cycles. According to the mathematical formula [18], the artificial mutational burden during 12 cycles of PCR with an error rate of ~10−4 per base per doubling was ~6.00×10−4/bp. However, the calculated artificial mutational burden was much higher than the 5.49×10−5/bp which we obtained by the experimental assessment of PCR error per base. This indicates that under the conditions of our experiment the true error rate of Taq DNA polymerase was lower than expected. Our method was designed to systematically detect precise somatic point mutations even when the degree of heteroplasmy was low. Therefore we believe that our method was particularly effective for studying somatic mutations.

The control region (D-loop) is known to be the most variable portion of the mDNA genome [17]. Our estimated aggregate somatic point mutational burden was of a similar size to the aggregate mutational burden found in the D-loop of two elderly persons (35~90×10−5) [45] and in brain mDNA (17.5±15.9×10−5 mutations per bp in all elderly subjects) [18]. As in our study, somatic transition mutations were also shown to be a major part of total somatic mutations [18, 45].

With regard to the reported somatic transition mutations caused by oxidation, oxidised C has been reported to cause C to T transition mutations from oxidative stress in E. coli [22]. These C to T or G to A transition mutations accounted for only a small percent of the somatic transition mutations detected in our study (10/25 in diabetic patients and 7/40 in control subjects, p=0.17 according to Fisher’s exact test). There were also no mutual relations between these transition mutations, HbA1c values and urinary 8-OHdG. Other studies have also used peripheral blood cells to examine the accumulation of A3243G mutations in mDNA of diabetic patients, quantifying its mutational burden as 5.2±0.3×10−5 mutations per bp [16]. Even allowing for the PCR error (~5.49×10−5/bp), this value is much lower than ours (31.66±6.63×10−5/bp). Hence, it is possible that nt 3,243 is not a hot point for mutation as stated in that report.

Concerning the correlation we found between the incidence of somatic transversion mutations in mDNA and HbA1c values, many reports have dealt with somatic mutations in mDNA. Most of them, however, have been on somatic transition mutations, deletions and insertions, whereas somatic transversion mutations have never been reported. In addition, only few documents exist on somatic point mutations associated with hyperglycaemia or HbA1c values. Our in vitro experiment using HL60 cells supported our findings on somatic transversion mutations in vivo.

In our study 8-OHdG in the urine of the diabetic patients was only a little higher than that in control subjects, although previous papers [7, 9, 13] showed considerably higher concentrations of urinary 8-OHdG in patients with diabetes mellitus. In addition, no correlation was found between urinary 8-OHdG and the transversion mutational burden. However, oxidative stress is reported to be increased in people with diabetes mellitus or hyperglycaemia [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], suggesting that the somatic transversion mutational burden of mDNA is associated with oxidative stress derived from hyperglycaemia, which is not reflected by urinary 8-OHdG.

Concentrations of urinary 8-OHdG in three diabetic patients who had microangiopathy or macroangiopathy or both, and also had 8-OHdG-associated mutations, were much higher than in the other subjects. These findings confirm the conventional idea [5, 9, 13, 15] that 8-OHdG is connected with diabetic complications. Mutations associated with 8-OHdG have not previously been shown in patients with diabetic complications. However, our study only detected three 8-OHdG-associated mutations, which did not correlate with HbA1c values.

It is possible that somatic mutations associated with hyperglycaemia occur not only in mDNA, but also in nDNA in very small amounts. In the light of many reports on diabetic complications in association with hyperglycaemia or HbA1c values [46, 47, 48], it is possible that such somatic transversion mutations also partly contribute to diabetic complications. Alternatively, common mechanisms could result in the complications and the somatic transversion mutations associated with hyperglycaemia. In conclusion, the increase detected by us in somatic transversion point mutations in mDNA from diabetic patients could be used as a novel biomarker.