Human Molecular Genetics

Figure 2. Schematic representation of haplotypes observed in 10 selected LHON subjects belonging to haplogroup D⁴²¹⁶₁₃₇₀₈ (4216+11251 +10398+12612+13708). D loop polymorphisms are given in italics.

DIDMOAD. We were able to demonstrate a statistically significant association between DIDMOAD and a distinct haplogroup (12 ). Six out of eight patients (75%) were associated with the 4216+11251+16126 lineage, of which five (63%) clustered in haplogroup D ⁴²¹⁶₄₉₁₇ (P < 0.01), with none in D ⁴²¹⁶_{1 3 7 0 8} (Table 2 ). Although one single DIDMOAD patient carried the polymorphisms 10398 and 12612, he presented as 13708-negative, forming a distinct branch [designated as the 13708(-) sub-cluster]. We also sequenced the two hypervariable D loop regions from all five DIDMOAD patients belonging to the D ⁴²¹⁶₄₉₁₇ haplogroup. Although members of a precisely defined haplogroup, they could be grouped in five different haplotypes, indicating again, as observed in the LHON group, that disease susceptibility lies within the coding region (haplogroup) and is not associated with any distinct D loop haplotype.MELAS. The haplogroup distribution of nine index patients with MELAS (eight patients with the 3243 mutation, one patient with the 3271 mutation) was investigated to answer the question regarding whether a distinct haplotype background or certain specific D loop variants might play a role in formation and fixation of disease mutations, as previously proposed for the D loop polymorphism 16189 (16 ,17 ). Here, we demonstrate that MELAS patients were similarly distributed among the distinct haplogroups, as were the controls (Table 2 ): two (22%) belonged to D ⁴²¹⁶₄₉₁₇, one to D ⁴²¹⁶_{1 3 7 0 8} (11%), another to lineage 12705 (11%) and the remaining five patients were members of haplogroup D^C (55%). D loop sequencing of all MELAS patients (data not shown) did not reveal any concentration in a distinct haplotype and showed that the MELAS mutations must have arisen multiple and independent times. Furthermore, we could not detect a higher frequency of the 16189 mutation, as reported by others (16 ), since only one MELAS patient (11%) harbored this polymorphism (controls 16%).SIDS. Nine children who had obviously died from SIDS were investigated for their association with haplogroups and haplotypes (Table 2 ). Three of the SIDS infants (33%) belonged to the 13708(-) side-branch of lineage 4216+11251, previously detected in one single DIDMOAD patient. This sub-cluster has not been found in our controls, since all control subjects carrying variants 4216+11251+12612+10398 also harbored the 13708 polymorphism. Two SIDS infants (22%) were members of the 12705 lineage, both also carrying a heteroplasmic C insertion at nt 574 characteristic for a sub-branch of D ^{1 2 7 0 5₁₀₃₉₈ in controls (3%). The children were obviously not related to each other as they were characterized by different D loop variants (data not shown). The heteroplasmic C insertion at nt 574 was also detected in a third SIDS infant, however on the background of lineage 12308+12372. Three SIDS children (33%) belonged to lineage DC. Although the clustering of SIDS in the 13708(-) side-branch as well as the increased frequency of the heteroplasmic polymorphism at nt 574 showed statistical significance (P < 0.005 and P < 0.01 respectively), the number of SIDS children belonging to the respective clusters was too small to provide clear-cut evidence for a disease-predisposing haplogroup.}

Table 3 Comparison of major German lineages and haplogroups with Caucasian data from the literature

Lineage	This study (SSCP/sequencing)			Torroni et al. (21)a (RFLP/D loop sequencing)			Richards et al. (20)a (D loop sequencing)
	Haplogroup	Frequency	Haplogroup	Frequency	Divergence (years)	Haplogroup	Frequency	Divergence (years)
7028 (-)	Dc	46%	H	40%	22 500-45 000	Group 1	29-77%	23 000
4216+11251	D 421613708	7%	J	3-15%	10 500-21 000	Group 2A	2-22%	23 000
	D 42164917	12%	T	6-22%	8276-10 909	Group 2B	0-13%	35 500
12308+12372	D 1230811467 + 16224	4%	K	8%	9750-19 500	Group 4	0-10%	17 500
	D 1230811467 + 3197	6%	U	10-16%	51 034-67 273	Group 5	0-21%	36 500
	D 1 2 3 0 84646	6%	U?			Group 5?
12705+16233	D 1270510398	4%	I	4%	10 500-21 000	Group 3A	5%	34 000
	D 127050195	6%	W,X?			Group 3?	4%	50 000
15904	D 15904072	4%	?			?

^aData are from Torroni et al. (21) and Richards et al. (20).

DISCUSSION

The mitochondrial genome had been fully sequenced 16 years ago (3 ) and many subsequent publications worldwide have revealed legions of polymorphic sites, as well as population-specific lineages, haplogroups and haplotypes. Nevertheless, a valid systematic classification remains elusive, although various authors have created their own nomenclature (12 ,18 -20 ). The rich variability within the D loop region compared with the relatively constant constellation within the gene regions provides useful criteria for phylogenetic and pathogenetic studies. Furthermore, the fact that distinct nucleotide positions have different tendencies to mutate has important implications for phylogenetic studies. We therefore suggest a classification of populations based on mtDNA gene variants. Our study demonstrates that the German population can be exactly characterized by a few haplogroups defined by common clusters of homoplasmic mtDNA mutations within the coding regions of the mitochondrial genome. Due to the lower mutability within the gene regions compared with the non-coding D loop, our haplogroup definition is affected with a much lower degree of uncertainty compared with the use of D loop sequences. For example, most of the identified coding region variants (~70%) occurred at least twice but showed strict haplogroup specificity, while ~30% of variants occurred only once in our control group. The only exception was the variant at nt 10398, which must have arisen multiple times during evolution as we found it in different haplogroups. In contrast, only ~25% of D loop polymorphisms could be classified as haplogroup-specific variants and the majority represented highly variable sites or very rare variants. Furthermore, the use of gene mutations leads to a more precise definition, allowing characterization of common extant sub-branches. Nevertheless, as we investigated our control population also with respect to the two hypervariable D loop segments we are able to compare our data with those described in the literature.

Comparing our results with those recently obtained by a combined RFLP/sequencing (19 ,21 ) and a HVS-I sequencing approach (20 ) lends support to the existence of the major German lineages and haplogroups (Table 3 ). In our control population a transition at nt 7028, corresponding to an AluI site at nt 7025 in Torroni et al. (19 ,21 ), separates the lineage D^C, which is 7028(-) in our nomenclature, from all other lineages. Furthermore, among 7028(+) subjects the major lineages 4216+11251, 12308+12373 and 12705 are all characterized by the transitional variant 073 in our study, a polymorphism also mentioned by Richards et al. (20 ) as a major separator between the consensus haplogroup (Cambridge Reference Sequence) and four longer branches. The main lineage 4216+11251, which is strongly correlated with the HVS-I variant 16126 (Fig. 1), diverges into haplogroups D ⁴²¹⁶₁₃₇₀₈ and D ⁴²¹⁶₄₉₁₇. These well-defined haplogroups were characterized by a number of homoplasmic genes mutations as well as D loop variants. They seem to be consistent with the haplogroups J and T in Torroni's grouping and 2A and 2B in Richard's nomenclature. Both haplogroups were reported to occur only in European populations. However, as mentioned by Richards et al. (20 ), the ancestral state of haplogroups 2A and 2B, namely the 16126 mutation alone, has only been found in the Middle East. Our second major lineage, 12308+12372, splits into three haplogroups designated D ¹²³⁰⁸_{11467 + 3197} , D ¹²³⁰⁸_{11467 + 16224} and D ¹²³⁰⁸₄₆₄₆, each characterized by several homoplasmic mitochondrial gene mutations. D ¹²³⁰⁸_{11467 + 16224} is probably consistent with Richard's group 4 and Torroni's group K, mainly because of the HVS-I variants 16224, 16311 and a HinfI restriction site gain at nt 12308. The D loop variant at nt 16270, which is associated with nt 3197, indicates that D ¹²³⁰⁸_{11467 + 3197} is consistent with groups 5 and U in Richard's (20 ) and Torroni's (21 ) studies respectively. Our third sub-cluster of lineage 12308+12372, haplogroup D ¹²³⁰⁸₄₆₄₆, although not equivalent to any specific haplogroup described in other studies, might also belong to Torroni's haplogroup U. Lineage 12705 branched into a 10398(-) and a 10398(+) haplogroup. The latter (D ¹²⁷⁰⁵₁₀₃₉₈) was defined by several further mitochondrial gene and D loop variants; it is likely to be consistent with Richard's group 3A. A sub-branch of D ¹²⁷⁰⁵₁₀₃₉₈ is defined by a heteroplasmic C insertion at nt 574, described by Torroni et al. (19 ) as a characteristic part of their haplogroup I. The lineage D^C in our study, called haplogroup H in Torroni's nomenclature, is equivalent to group 1 of Richards et al. (20 ). It is by far the most frequent group and comprises between 40 and 70% of the European population. However, as this group is defined by asubstraction of the other group-specific polymorphisms, it remains the most poorly defined group which, obviously, will be subdivided into further groups as our study continues. Examples of this are the rare haplogroups D^C₃₃₄₁, D^C₃₃₃₈, D^C₅₅₈₀, D^C₁₀₄₆₃, D^C₁₄₉₀₅ and D^C₄₃₃₆, which were 7028(-) and were found infrequently in our control population. D loop sequences of all 31 members of lineage D^C (Table 1 ) revealed 29 different haplotypes. Thus the existence of major and frequent haplogroups within this lineage seems unlikely.

The relative frequencies of the haplogroups described differ slightly within the central European area, as shown in Table 3 . Haplogroups D ⁴²¹⁶₁₃₇₀₈ and D ⁴²¹⁶₄₉₁₇ were found with a frequency of 7 and 12% respectively in our German control group. The distribution of both haplogroups, however, differs considerably across the Middle Eastand Europe (20 ,21 ,23 ,24 ). For instance, Torroni's group T (our haplogroup D ⁴²¹⁶₄₉₁₇ and Richards' group 2B) was not detectable in Turkey or Iceland (20 ), but was found with higher frequencies in the Middle East (12%), in Sardinia, Spain, Portugal (all 13%) (20 ,22 ) and, to the remarkably high degree of 75%, in a Ladin speaking isolate in Val Gardena in the northern part of Italy (24 ). It should be noted here that two of our DIDMOAD index patients belonging to haplogroup D ⁴²¹⁶₄₉₁₇ were of Turkish origin (12 ). Richard's haplogroup 2A (our haplogroup D ⁴²¹⁶₁₃₇₀₈) also varies widely within European populations, ranging from 2% in the Basques to 22% in Cornwall (20 ). Sub-clusters of lineage 12308+12372, averaging 7% in European countries (19 ,20 ), could not be found among Asians or Native Americans (20 ). Lineage 12705 is also found to a large degree in the Middle East (24%) and in Turkey (18%), but only to a much lesser extent in several European populations (20 ). We presume that part of the reported differences in the central European mitochondrial haplogroup distribution is due to uncertainties produced by hypervariable hotspots in the D loop sequences.

Recently Zischler et al. (6 ) described an integration of a mitochondrial D loop fragment into human chromosome 11, where it presumably evolves as a chromosomal pseudogene (`fossil') and, therefore, more slowly than its mitochondrial counterparts. The common ancestor of the insertion and the contemporary mitochondrial gene pool probably originated shortly before the genesis of a common ancestor of the current mitochondrial gene pool. When compared with this nuclear `fossil', the Cambridge reference sequence is found to contain 25 base exchanges. Of the mitochondrial lineages and haplogroups described here, only the 12308+12372 and 12705 lineages share several additional positions with the fossil sequence, namely an A at nt 16129, a C at nt 16249, a T at nt 16278, a C at nt 16311, a C at nt 16263 and a C at 16356. As pointed out by Torroni et al. (21 ), haplogroup U (our haplogroup D ¹²³⁰⁸₄₆₄₆) is the only one that Europeans share with Africans. Thus the 12308+12372 and, probably, 12705 lineages seem to be the oldest ones, originating possibly in Africa and expanding into the Middle East and Europe (21 ).

There also exists some uncertainty about the tree position and the value of the A -> G exchange at nt 10398. While Torroni et al. (19 ,21 ) suggested that the presence or absence of this DdeI site at nt 10394 separates all Caucasian mtDNAs into two major groups, our data suggest this polymorphism as being a hotspot occurring in different haplogroups. Otherwise, the haplogroups D ⁴²¹⁶₁₃₇₀₈ and D ⁴²¹⁶₄₉₁₇ would not share variants 4216+11251+16126 as an ancestral lineage and, furthermore, the major lineages 11251+4216+16126, 12308+12372 and 12705 must have arisen at least twice during evolution. A further possibility is that 10398 is indeed very ancient, but underwent back-mutation in certain haplogroups.

Most of the D loop variants characteristic of certain haplotypes have also been described in the skeleton network recently published by Richards et al. (20 ), summarizing the Caucasian HVS-I sequence data from DiRienzo and Wilson (22 ), Piercy et al. (23 ) and Pult et al. (25 ). Thus even at the level of the HVS-I fragment the Caucasian population can be identified by a limited number of variants. On the other hand, several positions within the D loop sequences represent mutational hotspots. Furthermore, most of the D loop positions investigated in the present and other studies (>80%) remain unaffected. Thus D loop positions seem not to be randomly affected by mutational events. Rather, a selection of mutational sites seems to be taking place. If so, this would have important implications for phylogenetic studies, because widely used methods such as maximum parsimony analysis (22 ,26 -28 ) and pairwise difference analysis (22 ,29 ) usually assign an equal weight to all nucleotide exchanges, as recently stated by Torroni et al. (21 ). Hence, for further phylogenetic studies the single D loop positions would have to be weighted individually.

The disease groups were chosen for several reasons. The LHON and MELAS groups consisted of patients routinely diagnosed in our laboratory. They represented typical maternally transmitted disorders with specific homoplasmic (most LHON patients) or heteroplasmic (all MELAS patients) mtDNA point mutations. Diagnostic studies 3 years ago on the LHON group had revealed that the bulk of the LHON patients were associated with a cluster of so-called secondary LHON mutations (9 ). By extending these studies to a group of healthy German subjects we found the frequency of haplogroup D ⁴²¹⁶₁₃₇₀₈ to be significantly higher in our LHON cases compared with controls (12 ,15 ), irrespective of the nature of the primary/intermediate LHON mutations. D loop sequencing of 10 LHON patients belonging to this haplogroup revealed that the 11778, 14484 and 15257 LHON mutations have arisen several times independently within haplogroup D ⁴²¹⁶_{1 3 7 0 8} (Fig. 2 ). The accumulation of LHON on this haplogroup background is unlikely to be the result of founder mutations, thus haplogroup D ⁴²¹⁶₁₃₇₀₈ seems prone to formation or expression of LHON mutations. The observation that another disorder exhibiting optic atrophy in combination with diabetes mellitus is concentrated in haplogroup D ⁴²¹⁶₄₉₁₇ (12 ) makes disease susceptibility at the level of lineage 4216+11251+16126 likely. DIDMOAD has been proposed to be a mitochondrial-mediated disease (30 ). However, it lacks maternal transmission and is thus unlikely to be commonly caused by severe mtDNA mutations (12 ,31 ,32 ). It has recently been mapped to chromosome 4p (33 ).

Children who have died from SIDS have been included in a mtDNA screening program and haplotype analysis for the following reasons. SIDS is the most common cause of death in infants from the first week to the first year of age, however, despite extensive global research the reasons and the pathomechanisms triggering the fatal event remain generally unknown. In a relatively high proportion SIDS is preceded by one or more so-called `near miss for SIDS' or `apparent life threatening events (ALTE)' which are characterized by respiratory arrest, cyanosis, increased temperature with massive sweating, etc. Obviously ALTE symptoms can be related to alterations in coupling of the respiratory chain and the oxidative phosphorylation apparatus. Furthermore, certain mtDNA mutations might affect coupling (2 ) and this could have pronounced effects, especially in the mitochondria-rich brown adipose tissue, which regulates thermogenesis in the newborn during the first year of life. Although a detailed mtDNA screening for disease mutation by our group (unpublished results) did not reveal any clear-cut evidence for SIDS-causing mtDNA mutations, our haplogroup analysis showed a different distribution compared with our control population. First, the SIDS infants accumulated in a side-branch of 4216+11251, characterized by additional variants at nt 10398 and 12612, a haplogroup not found in 67 German controls. Second, 3/9 SIDS children carried a heteroplasmic C insertion at nt 574 while only two control subjects (3%) carried this polymorphism, in both cases on a haplogroup D ^{1 2 7 0 5}_{0 1 9 5} background. These unstable C stretches in the non-coding region have been previously discussed as predisposing to somatic mutations (19 ), as also suggested for unstable stretches of C residues around nt 16189 (34 ). Whether this variant plays a role in susceptibility for SIDS needs further investigation.

The association of different disorders with certain mitochondrial haplogroups seems to us the most important result of our study. If this can be confirmed by further studies it will have unforeseen consequences and implications for genetic research in general. If the idea of disease-susceptibile haplogroups holds true, populations with, for example, higher frequencies of haplogroups D ⁴²¹⁶_{1 3 7 0 8} and D ⁴²¹⁶₄₉₁₇ should show an increase in the penetrance of distinct disorders. On the other hand, populations with higher frequency of haplogroup D^C should be relatively protected from such diseases.

Is there any rational explanation for the association of certain mtDNA haplogroups and diseases? mtDNA encodes several peptide subunits necessary for optimal functioning of the respiratory chain and the oxidative phosphorylation apparatus in the mitochondrion, responsible for the essential production of ATP in all organs and cells, especially in those with high energy demand such as brain, heart and skeletal muscle. It is therefore not surprising that single mtDNA gene mutations cause functional deficiencies in the corresponding gene products, thus leading to insufficient function of the energy-providing machinery and, ultimately, to monogenic mitochondrial diseases of the brain, heart, skeletal muscle and neuroendocrine system (2 ). To our knowledge there are only a few publications dealing with the role of polymorphic mitochondrial genetic backgrounds in cellular function and individual cellular fitness. In 1991 Dionne et al. (35 ) correlated the mtDNA RFLP pattern of atheletes with maximal oxygen uptake and response to endurance training and found that distinct sequence variation clusters were associated with different levels of physical fitness. Recently Vergani et al. (36 ) showed that the LHON 11778 mutation leads to a deficiency in ATP production, especially when associated with the polymorphic site at nt 4216. This study supplied, for the first time, a functional basis for the assumption that lineage 4216+11251 might increase the risk of expression and penetrance of LHON and, possibly, other disorders. Thus the disease susceptibility of certain haplogroups might be caused by relative differences in energy production due to clusters of synergistically working point mutations, as already proposed by others (10 ,11 ,13 ). Furthermore, certain haplogroups (or specific mtDNA variants) might be more prone to the formation of disease-causing mutations; this possibility has been discussed for the heteroplasmic C stretches at nt 574 (19 ) which were detected with high frequency in our SIDS group. A third possibility is that different mtDNA mutation clusters might generate different amounts of oxygen radicals, with their respective consequences on DNA, protein and lipid structures. However, up to now evidence for any such explanation is weak; obviously, further studies are needed.

MATERIALS AND METHODS