Mitochondrial D-loop Mutations As Clonal Markers in Multicentric Hepatocellular Carcinoma and Plasma¹

HCC³ is one of the most common malignancies worldwide and is generally associated with a poor prognosis (1). The disease is highly prevalent in Asia but relatively rare in developed countries, although the incidence is increasing in both the United States and Japan (2). Many environmental factors, including aflatoxins, alcoholism, and hemochromatosis, as well as a susceptible genetic background, have been implicated in the development and progression of HCC (3). However, chronic HBV and HCV infection are by far the best documented factors associated with the progression of chronic hepatitis to cirrhosis and eventually to HCC (4).

Although other therapeutic approaches have been introduced to treat HCC (5), surgery remains the treatment of choice. The progression and outcome of truly relapsed HCC compared with a second primary tumor are distinct (6), and thus, clonal analysis of the initial and recurrent HCC is of clinical significance. Although several studies have addressed HCC clonal determination (7, 8, 9), reported molecular methods are still of limited use clinically or are technically challenging.

Recently, we and others (10, 11) identified a high frequency of missense and frameshift mutations in mtDNA from primary human neoplasms. These mutations were most frequently found in the noncoding D-loop region (11), especially in a polymorphic cytosine mononucleotide repeat between nucleotides 303 and 315. Although occasional defects of complex III (ubiquinone-cytochrome c-oxidoreductase) and IV (cytochrome c-oxidase) within the respiratory chain have been detected in both normal and cirrhotic livers, (12) the existence of mtDNA alterations in HCC was not generally recognized. More recently, one group reported frequent mtDNA mutations in HCC, as well as in noncancerous tissue (13).

We report frequent mutations in the control region of the mitochondrial genome in tissue samples from HCC. The high frequency of mutations within the control region of the mtDNA provided a tool to determine the clonal origin of multiple HCCs in individual patients. In addition, mutant mtDNA was detected in matched plasma samples using sensitive techniques and might one day prove useful for management of this disease.

HCC and adjacent nontumorous liver tissues from 19 patients were surgically resected in Nagoya University Hospital between 1989 and 2000. Thirteen of 19 patients had synchronous multiple HCC, which were resected during their first operation. In the other 6 patients, hepatic tumors reappeared between 10 months and 5 years after the first curative operation and were resected by a surgical approach. Serological markers for HBV and HCV infections and α-fetoprotein levels were recorded.

All specimens were immediately fresh frozen after resection and stored at −80°C. Serial 10-μm sections were cut with a microtome. After staining with H&E for histological examination, the other sections were used for DNA extraction. With the exception of small areas comprised of fatty and necrotic cells, nearly all sections were comprised of >90% neoplastic cells. In 14 of 19 cases, we collected plasma samples before treatment, and these were stored at −80°C until extracted for DNA. Normal, tumor, and plasma DNA were prepared as described previously (14).

To avoid amplification of nuclear-encoded pseudogenes (15), mtDNA fragments of >2.5 kb in length containing the D-loop region were amplified using the following primers F47: 5′-CGCACGGACTACAACCACGAC-3′ (forward) and R15: 5′-TGTGGGGGGTGTCTTTGGGG-3′ (reverse). In a PCR buffer containing 5% DMSO, 100 ng of genomic DNA were subjected to the step-down PCR protocol: initial denaturing step at 94°C for 5 min followed by 94°C for 45 s, 64°C for 1 min and 15 s, and 70°C for 3 min for two cycles; 94°C for 45 s, 61°C for 1 min and 15 s, and 70°C for 3 min for two cycles; 94°C for 45 s, 58°C for 1 min and 15 s, and 70°C for 3 min for 30 cycles; and a final extension at 70°C for 5 min. PCR products were gel purified using a Qiagen gel extraction Kit (Qiagen, Inc., Valencia, CA). In each experiment, DNA extracted from a mtDNA-negative cell line (ρ⁰) was included as a negative control (16).

Amplified fragments containing the mitochondrial D-loop region were sequenced using (γ-³³P) ATP 5′ end-labeled sequencing primers and an AmpliCycle sequencing kit (Perkin-Elmer and Roche Molecular Systems, Inc., Brachburg, NJ) under the following cycle conditions: 95°C for 30 s, 52°C for 1 min, and 70°C for 1 min for 30 cycles. The sequence of the primers was, F44: 5′-GTCTTTAACTCCACCATTAG-3′, F55: 5-ACCGTACATAGCACATTACA-3′, F56: 5′-ATCACGATGGATCACAGGT-3′, and F57: 5′-CGTTCAATATTACAGGCGAAC-3′. The sequenced products were analyzed on a denaturing 6% polyacrylamide gel.

A simple PCR assay, based on the amplification of a 104-bp product that included the C-mononucleotide repeat, was performed for rapid detection of C-tract deletion/insertion mutations. In PCR buffer containing [³²P] dCTP, 100 ng of genomic DNA were subjected to the following PCR protocol: initial denaturing at 94°C for 3 min followed by 94°C for 20 s, 55°C for 20 s, and 70°C for 15 s for 35 cycles and a final extension at 70°C for 5 min using primer sets of C-tract F: 5′-TCTGACCAGCCACTTTCCA-3 and C-tract R: 5′-GGGTTTGGCAGAGATGTGT-3′. PCR products were analyzed on a denatured 6% polyacrylamide gel and processed by autoradiography.

Because of the low concentration of DNA in the matched plasma samples, mtDNA mutations were not detected using sequence analysis alone, and a more sensitive oligonucleotide mismatch ligation assay was performed (17). We applied this method to matched plasma DNA to 4 of 5 cases, in which we detected primary tumor missense mutations. In the cases with C-tract deletion/insertion mutations, we could not perform the assay because of a problem with oligo design. Fragments containing the mitochondrial D-loop region were PCR amplified and extracted using the method described previously. For each missense mutation, discriminating oligonucleotides that contained the mutated base at their 3′ end were designed (position: 72, 5′-CTGGGGGGC-3′ for patient 1, position: 16278, 5′-CTAGGATAC-3′ for patient 14, position: 16390, 5′-CAGATAGGA-3′ for patient 16, and position: 70, 5′-GTCTGGGGA-3′ for patient 19). An immediately adjacent 3′ oligonucleotide linker, along with the discriminating oligonucleotide (5′-ATGCACGCG-3′ for patient 1, 5′-CAACAAACC-3′ for patient 14, 5′-GTCCCTTGA-3′ for patient 16, and 5′-GTATGCACG-3′ for patient 19) was used as a substrate for the ligation reaction. In case 16, a blocking oligo (5′-TAGGGGTCCC-3′) was used for decreasing the background. Discriminating oligonucleotides (250 ng) were mixed with the PCR amplified fragment and 40 ng of the γ-[³²P] end-labeled 3′ oligonucleotide linker. The reactions were incubated at 37°C for 1 h in the presence of T4 DNA ligase (Life Technologies, Inc., Grand Island, NY), analyzed on denatured 12% polyacrylamide gels, and processed by autoradiography.

The d-loop region of the mitochondrial genome was analyzed in all tumors from 19 patients with multiple HCCs and matching nontumorous tissue by direct sequencing. Representative results of the sequence analysis in normal tissue and paired tumor is shown in Fig. 1. In case 1, all three tumors harbored a T to C transition. Detailed results of the mitochondrial sequence analysis are summarized in Table 1.

Thirteen of 19 patients (68%) harbored a D-loop mtDNA mutation in at least one tumor. Because many mutations were in the D-loop, we also tested all samples with a simple PCR fragment length assay. This assay is based on the amplification of a 104-bp product that includes the C-mononucleotide repeat. Comparisons between direct sequence analysis and the PCR assay are shown in Fig. 2. We observed C-tract deletion/insertion mutations in 8 of 19 cases (42%) and five other missense and deletion/insertion mutations in an additional 5 cases (26%). Eight cases (numbers 5, 7, 12–16, and 19) revealed mutations that were found in only one of the tumors tested (Fig. 2, a and b). One case (number 17) revealed the same mutation in two of three tumors, whereas all tumors from cases 1 and 10 shared identical mutations. Finally, in cases 2 and 4, each tumor tested harbored a different mutation. All of the somatic mutations detected were homoplasmic or near homoplasmic, confirming results from previous studies in other tumor types (10, 11).

The results of our molecular analysis were compared with more standard clinical and pathological criteria in Table 1 (“Clinical Diagnosis”). These criteria are based on pathological findings, including tumor size, location of the lesion (e.g., right lobe or left lobe), and interval of occurrence. In some cases (e.g., cases 1 and 17), molecular analysis identified the presence of an identical mutation confirming the clinical diagnosis of metastases. In other cases (e.g., case 10), molecular analysis identified the presence of an identical mutation, whereas the clinical diagnosis was indeterminate. In some cases (e.g., case 4), distinct mutations confirmed the independent origin of the tumors by clinical analysis. In at least 1 case (case 2), independent mutations argued against the clinical diagnosis of metastases.

In addition to the somatic mutations shown above, we identified 38 polymorphisms, of which 5 are apparently newly described (Table 2). There was no significant association between HBV status, HCV status, age, or gender and the occurrence of mtDNA mutations (by Fisher’s exact test).

Plasma DNA was available preoperatively from 10 of 13 cases where a mtDNA mutation was detected in a primary HCC. We applied a more sensitive oligonucleotide mismatch ligation assay to detect missense mutations, and in 3 cases, we confirmed the mutant sequence in tumor DNA and documented a negative signal in the corresponding normal tissue. We then demonstrated that mutant mtDNA could be detected in the plasma DNA samples from all 3 of these cases (cases 1, 14, and 19; Fig. 3,a). In the remaining case with a missense mutation (case 16), there were weak background bands in both normal and plasma DNA despite the use of a blocking oligo. We applied the simple PCR fragment length detection assay to identify mutant mtDNA in 6 cases with C-tract deletion/insertion mutations, and in 5 of these 6 cases, we detected the identical C-tract deletion/insertion mutations in matched plasma DNA (Fig. 3 b). Overall, we detected mutant mtDNA in 8 of 10 plasma DNA samples. In all cases, the mtDNA mutant signal was much weaker than the tumor (range 1:200–1:500), reflecting significant dilution in or lack of mtDNA access to the plasma.

Recent advances in our understanding of the structure and function of mitochondria have led to the recognition that both inherited and acquired mitochondrial dysfunction may be responsible for diseases affecting the liver and other organ systems (18). Previously, two types of liver disease with related mitochondrial alterations were reported (18). Primary mitochondrial hepatopathies are conditions in which inherited defects in structure or function of the mitochondria are present, leading to hepatic failure or chronic liver dysfunction in affected children. Secondary mitochondrial hepatopathies are conditions in which the mitochondria are major targets during liver injury from another cause, such as metal overload, certain drugs and toxins, alcoholic liver injury, and conditions of oxidative stress. We found frequent mitochondrial mutations in HCC tissue. Mitochondria that undergo rapid replication in chronic hepatitis or in HCC may acquire and accumulate DNA damage more readily than those maintained under resting condition (13). These somatic mitochondrial mutations in cirrhotic liver tissue or in HCC may be best classified with the secondary mitochondrial hepatopathies.

Recently, we and others (10, 11) identified a high frequency of missense and frameshift mutations in the mtDNA from primary human neoplasms. In some cases, these mutations could lead to abnormal metabolic and apoptotic processes in neoplastic cells (e.g., many coding mutations were confined to respiratory complex I; Ref. 11). However, many identified somatic mutations are silent or occur in noncoding regions, and their significance is unclear (see further discussion below). The noncoding D-loop was found to be a mutational hot spot in bladder, lung, and head and neck neoplasms (11). Mutations in this region may alter the function of the D-loop, as this represents a regulatory site for both replication and expression of the mitochondrial genome. We found acquired somatic mutations in 13 of 19 cases not present in the matched nontumorous lesions in the D-loop region. All of the somatic mutations detected were homoplasmic, confirming results from previous studies in which the mutated mtDNA apparently becomes dominant in tumor cells (10).

Remarkably, eight of the mtDNA alterations were deletion/insertion mutations in a polymorphic mononucleotide repeat sequence (CCCCCCCTCCCCC) between nucleotides 303 and 315. We recently detected these alterations in 24% of all tumors tested (n = 161) and in every tumor type analyzed except ovary and prostate (19). The biological reason for these alterations is unclear, but this region is the start site of replication of the closed, circular mitochondrial genome (20). Replication of mtDNA begins with the synthesis of the heavy strand (H strand) using primer RNA, and the 3′ termini of primer RNA were mapped to CSBs I-III (21, 22). We found frequent deletions or insertions only in the CSB II region in HCC samples but not in other CSBs, suggesting some functional relevance. Kang et al. (23) reported that the RNA-DNA hybrid at CSB II was more stable than those at the other CSBs in vivo. Interestingly, we did observe a 53- and 10-bp deletion in this area in two different primary HCCs from 1 case (case 2). The identification of larger deletions in this region serves as an impetus for further research on the mechanisms of late replication and processing of mtDNA in cancer.

Previous studies have used the patterns of HBV DNA integration in the cancer genome or selective chromosome inactivation for clonal analysis (7, 8). However, these methods can be applied only to HBV carriers or female patients, respectively. Because most patients with HCC are male and an increasing number of cases are associated with HCV, there is a need for clonal molecular markers. Recently, a comparative genomic hybridization study documented chromosomal aberration profile as a unique fingerprint in HCC (9). Although comparative genomic hybridization profile comparisons can help establish clonality in HCC, this method is limited and requires special analysis systems. Tumor-specific molecular abnormalities have been investigated as a diagnostic tool in HCC. Frequent mutations of p53 were used for determination of clonality of multiple HCC lesions (24). However, the frequency of p53 mutations in HCCs may be low because of complementary viral inactivation (25, 26), and as such, alterations of the p53 gene are generally found only in advanced stage (27).

Clearly, tumors studied that have the identical mtDNA mutation are of the same origin. Cases 1, 10, tumor 2 of case 17, and those that display different mutations are likely to have arisen from separate independent foci (cases 2 and 4). If the primary tumor has a mutation and the second tumor reveals no mutation, they should be considered independent lesions (cases 12, 13, 15, 16, 19, and tumor 3 of case 17). It is difficult to determine clonality when the primary tumor does not reveal any type of mutation and the second tumor has a mutation, because mtDNA mutations may occur at any time during progression and may thus not always precede all waves of clonal outgrowth (cases 5 and 7). However, a tumor which has a different mutation compared with the primary lesion has at least had clonal divergence, although it may represent a metastasis from a clinical point of view.

In our study, the clinical diagnosis of metastasis or independent lesions was based on the pathological findings that include tumor size, location of the lesion, and interval of occurrence (28), whereas in some cases, we could not determine their clonal relationship because we lacked more comprehensive markers. Although some discrepancies occurred between mtDNA mutation status and clinical diagnosis, mtDNA mutations can help establish clonal relationship and select an appropriate treatment strategy for new primary tumors or recurrent cancers. This molecular approach requires minimal DNA from each tumor and can be used to analyze routine needle biopsies.

We also demonstrated that mutant mtDNA could be detected in plasma samples from most patients with HCC. These results are probably because of the homoplasmic nature of the somatic mtDNA mutations within each cell, providing an unprecedented detection advantage because of the high copy number of mutant DNA. The low absolute abundance of mutant mtDNA detected in plasma suggests that mtDNA may not be readily introduced into or survive long in the circulation. This approach may prove useful as part of a panel of noninvasive DNA-based tests for the detection and monitoring of HCC after additional study.

We have shown that control region mtDNA mutations are a frequent event in HCC and that these mutations provide significant information in determining the clonality of multiple HCCs. These mutations may serve as sensitive markers to distinguish metastatic disease from the occurrence of multiple independent primary tumors.

We thank James Engles and Henning Usadel for helpful discussion. We also thank Robert Yochem and Robin Brewster for administrative support in preparing this study.