Evaluating sequence-derived mtDNA length heteroplasmy by amplicon size analysis.

Length heteroplasmy (LH) in mitochondrial (mt)DNA is usually observed in homopolymeric tracts and manifest as mixture of various length variants. The generally used difference-coded annotation to report mtDNA haplotypes does not express the degree of LH variation present in a sample, even more so, it is sometimes difficult to establish which length variants are present and clearly distinguishable from background noise. It has therefore become routine practice for some researchers to call the dominant type, the "major molecule", which represents the LH variant that is most abundant in a DNA extract. In the majority of cases a clear single dominant variant can be identified. However, in some samples this interpretation is difficult, i.e. when (almost) equally quantitative LH variants are present or when multiple sequencing primers result in the presentation of different dominant types. To better understand those cases we designed amplicon sizing assays for the five most relevant LH regions in the mtDNA control region (around ntps 16,189, 310, 460, 573, and the AC-repeat between 514 and 524) to determine the ratio of the LH variants by fluorescence based amplicon sizing assays. For difficult LH constellations derived by Sanger sequencing (with Big Dye terminators) these assays mostly gave clear and unambiguous results. In the vast majority of cases we found agreement between the results of the sequence and amplicon analyses and propose this alternative method in difficult cases.

Introduction

Length heteroplasmy (LH) describes the co-existence of at least two populations of mitochondrial (mt)DNA molecules in a DNA extract that differ in the number of nucleotides and therefore in their length. This generally leads to difficulties when Sanger sequencing data are to be interpreted, as the individual nucleotide sequences overlay after LH regions and appear as blurred pattern. While LH is usually observed in homopolymeric tracts when the number of identical adjacent nucleotides is greater than eight (for C-tracts), it has also been found in the AC-repeat and less often in shorter length variants of the control region (CR) [1-5]. In rare cases LH has even been observed in non-repetitive sequences [6]. The human mtDNA CR comprises five regions where LH occurs regularly and it is most frequently observed in the polycytosine tracts of the two hypervariable regions, HVS-I and HVS-II (Fig. 1) [3,5,7-10]. More than half the samples of a systematic study on 5015 individuals exhibited length heteroplasmy somewhere in the control region [11], whereas typical frequencies for LH to appear in polycytosine tracts are 12–20%, 45%, and 3–5% for HVS-I, II and III, respectively [11-15]. Note that the frequency of LH in HVS-I can also reach much higher levels exceeding 50% in populations where the transition T16189C that leads to an uninterrupted polycytosine tract represents a common and haplogroup (hg)-specific signature mutation (e.g. hg B in East Asia and Native America [11]).

Fig. 1

Schematic representation of the amplicon location for the analysis of the five targeted length heteroplasmic regions within the mtDNA control region. X-axis sized in bp. The hypervariable regions range from positions 16,024–16,365 (HVS-I), 73–340 (HVS-II) and 438–574 (HVS-III), respectively.

The biological relevance of LH is not yet fully understood. Tissues of an individual may display LH to various extents [16], especially when they go through a tight bottleneck in development, such as observed in mitochondria of hair shafts. The interpretation of LH in population genetics is usually disregarded. Forensic science sometimes also does not take LH into account when interpreting evidence (exclusion scenarios) and when the rarity of a sequence is determined in a database search. Since quantitative information (on the mixture of LH variants) cannot be captured by the difference-coded annotation of a haplotype (with respect to the revised Cambridge Reference Sequence, rCRS [17]) the number of detectable LH variants is not described properly. Also, it is sometimes technically difficult to precisely distinguish LH variants from background noise, especially since their depiction by the raw data may vary depending on the sequencing chemistry used. PCR primers, Taq and other polymerases, sequencing primers and the version of the sequencing kit do have an effect on the depiction of the raw data by influencing peak heights in a sequence-dependent context. An extreme effect is demonstrated by an example in [18].

As a result, it has therefore become routine practice to call the dominant type (major molecule) of the LH variants present in a sample [19]. This standard encoding identifies the most abundant of the length variants, which can clearly be determined in the vast majority of cases. This can be achieved by determining the highest signal in a peak pattern that a singular (non-repetitive) nucleotide (here referred to as “LH marker”) produces as a consequence of the different length variants present in a sample, e.g. at positions 16196G, 316G or 577G. In many cases this peak pattern displays a bell-shaped phenotype that shows the dominant LH variant in the mid-range size.

However, there are examples where this interpretation is not straight forward as equally quantitative LH variants may be present that then result in identical signal heights of the LH marker. In these and other complicated cases we applied the idea of investigating LH with fluorescence-based amplicon size detection as demonstrated in principle earlier [3,4,20] by expanding the assays to cover all five major LH regions in the mtDNA CR (Fig. 1).

Materials and methods

Generation of synthetic amplicons for mixture analysis

For each LH region two corresponding nucleotide sequences differing in length by one (polyC-tracts) and two bases (AC-repeat) were synthesized commercially and PAGE purified grade (Microsynth, Balgach, Switzerland). They were used to evaluate the detection of mixtures at ratios of 1:1, 1:0.95, 1:0.9, 1:0.85, and 1:0.80 (number of replicates ≥5, Table S1). The shorter variant was defined as standard (100%, component A) and the longer variant was added as equal or minor contribution (component B).

DNA extraction

Blood, buccal swab and autopsy tissue samples were extracted using the BioRobot M48 workstation (Qiagen, Hilden, Germany), and the MagAttract DNA M48, Mini Kit (192) according to the manufacturer's protocol and the Qiasoft M Operating System software (ver.2.0E001).

Quantification assay

MtDNA genome equivalents were determined in the extracts based on 143 bp mtDNA amplicons using a real-time quantitation PCR (rtqPCR) assay according to [21]. Concentration and purity of the synthetic DNA fragments were assessed using the NanoDrop ND-1000 Spectrophotometer (PEQLAB Biotechnologie, Erlangen, Germany) with standard instrument settings as recommended by the manufacturer. Extraction blanks as well as negative and positive controls were carried through the laboratory processes.

PCR and amplicon sizing analysis

The five targeted LH regions were amplified in singleplex PCR reactions in a total reaction volume of 20 μl including 1× PCR buffer, 1 mM MgSO4, 200 μM each dNTP, 0.4 U KOD Polymerase (KOD Hot Start DNA Polymerase, Novagen-Merck KGaA, Darmstadt, Germany), 0.25 mg/ml BSA (Serva, Heidelberg, Germany), 0.6% Trehalose, 150 nM each primer (see Table S2). Primers were designed using Oligo Analyzer from the homepage of Integrated DNA Technology (http://eu.idtdna.com/home/home.aspx). Unambiguous and reproducible results were obtained with 500 mtDNA Genome Equivalents (mtGE) for the AC repeat region and C-tract around ntp 460, 100 mtGE for the C-tract around ntp 573 and 2000 mtGE for the C-tracts surrounding ntps 16189 and 310. Thermal cycling was performed on a Gene Amp PCR System 9700 (Applied Biosystems, AB, Foster City, CA, USA) comprising an initial denaturation step at 94 °C for 2 min, 30 cycles of 95 °C for 15 s, 56 °C for 7 s. Aliquots of 1.5 μl of the amplification products were combined with 20 μl deionized formamide and 0.4 μl internal size standard (Genescan-500 LIZ, AB), heat denatured at 95 °C for 3 min, snap-cooled on ice, and subjected to capillary electrophoresis on an ABI Prism 3100 Genetic Analyzer using POP 6, 36 cm capillary arrays and default instrument settings. The data were analyzed using GeneScan Analysis version 3.7 and Genotyper version 3.6 (both AB).

Results and discussion

The interpretation of LH and the assignment of dominant types using direct Sanger sequencing are straight forward in the vast majority of mtDNA haplotypes encountered in population and forensic genetics. In some instances however, two or even more length variants may be present in similar quantity, which then makes the identification of a single dominant type difficult or even impossible. Also, it has been observed that the location of the sequencing primer has an effect on the depiction of the dominant type as far as the LH markers are concerned. Therefore multiple redundant sequencing reactions may not result in the same dominant type when different primers were applied. In these cases alternative methods may be helpful to solve ambiguities. For this purpose we designed fluorescent-based amplicon sizing assays that target the five most relevant LH regions within the mtDNA CR (around ntps 16,189, 310, 460, 573, and the AC-repeat between 514 and 524). The PCR setup was performed using the proof reading KOD Polymerase to avoid non-template 3′ overhangs that would otherwise result in artificially (by PCR) created length variation. The primers were selected in a way that each amplicon can be co-amplified with any other LH region in duplex format, if more than one region needs to be investigated in a sample (see Fig. 2).

Fig. 2

Examples of amplification plots for the targeted length variant regions with standard samples that do not display heteroplasmy. X-axis sized in bp.

Mixture studies

Since the most difficult constellations for LH interpretation are (almost) even mixtures of different length variants the performance of the five assays was evaluated using synthetically produced length variants at defined mixture ratios of 1:1, 1:0.95, 1:0.9, 1:0.85, and 1:0.80. The synthesized nucleotide sequences differed by only one additional C-residue for testing C-tract regions and in one AC unit for the AC repeat region. For all these mixture samples the shorter fragment (A) was defined as dominant variant (100%) and consequently the longer fragment (except for the 1:1 mixture) always constituted the minor component (B). The latter was always clearly and even visually discernable from the major peak up to mixture ratios of 1:0.95 for all LH regions (Fig. S1a–g), which was also confirmed by comparing mean peak height values between runs. Note that the individual mixtures have been prepared de novo independently prior to PCR, which may have caused the relatively high standard deviation values due to pipetting error. This is also visualized in the graphs that display the results of the mixture ratio study for all five LH regions, where the dotted lines represent the theoretical mixture values and the straight lines correspond to the regression curves of the measured mixture ratios. These results demonstrate that the amplicon sizing assays are suitable for determining the dominant variant in LH mixtures up to a ratio of 95%.

Additionally, LH mixtures were produced using quantified DNA extracts from blood samples of individuals that were known from prior sequence analysis to vary in length variant regions 310 and the AC repeat region. Thus, mixture experiments with these samples as outlined above created a picture that resembles LH. The amplicon sizing assays were also conducted multiple times (n ≥ 5) and confirmed the findings of the analyses performed with the synthetic DNA fragments (Fig. S1f and g).

Species specificity

DNA samples from 14 different vertebrate species were amplified with each amplicon sizing assay (Table S3). The amplicon sizes obtained for the human sample correspond to the rCRS variant. The 51 bp fragment observed for the amplicon surrounding the C-tract around ntp 573 in the human sample was also found in the gorilla, orang-utan and chimpanzee where the amplicon was slightly larger (52 bp). All other samples brought a 10 bp shorter amplicon, except for the hare (Lepus europaeus), which gave no amplification product at all. Interestingly, gorilla and orang-utan samples gave both amplification products. For the AC repeat region only the chimpanzee sample showed an amplification product within the expected amplicon size for humans. Patterns of multiple amplicons with different sizes were found for other animals including pig, cat, sheep, rat, and falcon. The three other amplicons did not reveal amplification products in the animal samples.

Population study and GEDNAP trial example

In order to compare results of direct Sanger sequencing and amplicon sizing we searched the mtDNA datasets of [22-24] for instances of LH in the electropherograms and typed 48 affected samples with the amplicon sizing assay. Generally, we found good agreement between the results in terms of identifying the dominant type. In some cases the assignment of a dominant type by Sanger sequencing was not unambiguously possible (Fig. S2a and b). Therefore it was interesting how the amplicon sizing assays would perform here. The example presented in Fig. S2a shows LH in the C-tract around ntp 573 consisting of at least five detectable length variants. The sequencing electropherogram suggests the presence of two almost identical dominant variants with a slight predominance of the shorter of the two, which is represented by 4 C-insertions (10 C residues in total), as determined by the height of the LH marker peak at position 577G (confirmed by the sequencing electropherogram of the reverse strand). The amplicon sizing assay revealed at least six LH variants in that sample but confirmed the relative dominance of the molecule represented by 10 C residues, even in a clearer way as the results of direct Sanger sequencing suggest. In another example (Fig. S2b) the HVS-I C-tract around ntp 16189 was difficult to interpret by sequence raw data. The sequencing electropherograms show three G residues in the LH marker 16196 with nearly identical peak heights indicating three almost equally dominant types, two of which (the longer ones) gave slightly larger peak heights compared to the shorter variant. The amplicon sizing assay brought a clear dominant type (16193.1C) and relative ratios of 0.89 and 0.93 for the shorter and longer variants, respectively.

The third example (GEDNAP trial 38 sample D; concept see [25]) presents a difficult LH situation in the C-tract around ntp 310. Two almost equally abundant length variants gave different dominant types depending on the applied sequencing primers (Fig. S2c). The majority (22 out of 30) of the GEDNAP participants reported 309.2C as dominant type, while 7 opted for 309.1C (one laboratory used a different nomenclature). The amplicon sizing assay gave a clearer picture with a mixture ratio of about 1:0.75 in favour of the 309.2C type (Fig. S2c).

Tissue study

From decomposed autopsy material we received nine different tissue samples (buccal swab, blood, liver, lung, kidney, spleen, heart, skin, psoas) from two individuals, where one of them did not exhibit LH in any of the investigated tissues and regions. The other individual showed LH in the C-tract around ntp 310 and in the AC repeat region. As a general observation by Sanger sequencing the two heteroplasmic regions showed comparable mixtures in each tissue. However, in the AC repeat region of the blood and the liver the electropherograms brought a slightly different distribution of the components (Fig. S3a). In both samples the dominant type was manifest as (AC)7 while the minor component included 5 AC repeat units (equivalent to the rCRS). The relative proportion of the minor component was higher in the liver tissue compared to the blood sample (by approx. 15%). In the blood and saliva samples the C-tract around ntp 310 exhibited three dominant variants with 12, 13 and 14 C residues in the sequence electropherograms, respectively (Fig. S3b). These findings were also confirmed with the reverse sequencing strands. The amplicon sizing approach identified the 13 C molecule (rCRS) as dominant type in the blood sample with a ratio of 1:0.84:0.70 compared to the 14 C and 12 C molecules. In the buccal swab this ratio was shifted to 1:1:0.60, resulting in two indistinguishable dominant types of 12 C and 13 C molecules.

Conclusions

The interpretation of mtDNA LH has not yet been rigorously discussed in the forensic community, nor is there general consent in a harmonized way of reporting LH. One main reason may be that the experimental approach is difficult and sometimes leading to ambiguous results as also demonstrated in our study. Our data suggest that fluorescence-based amplicon sizing is more robust for quantitative determination of dominant LH variants than direct Sanger sequencing with Big Dye terminators. Also, amplicon sizing seems to be more sensitive for the determination of the dominant type, especially in mixtures that appear (almost) even in sequence electropherograms. In practical work those LH constellations are most difficult to interprete as a number of factors seem to influence the interpretation including the selection of sequencing primers. The amplicon assay then offers an independent approach to gain more data that may be useful to derive an unambiguous decision. In all instances did we observe that a clear dominant type derived by sequence raw data was confirmed by amplicon analysis, which confirms the suitability of the method. The proposed tools may serve as helpful alternative in cases where a dominant type cannot be retrieved from sequence raw data.