Crypt-Restricted Loss and Decreased Protein Expression of Cytochrome c Oxidase Subunit I as Potential Hypothesis-Driven Biomarkers of Colon Cancer Risk

There is an increasing demand for the development of intermediate biomarkers to assess colon cancer risk (1, 2). It has been a long-standing goal of our laboratory to establish such a reliable biomarker that would evaluate colon cancer risk on an individual patient basis. One specific biomarker that we have found is an increase in apoptosis resistance, assessed using “stressed” colonic mucosa in an ex vivo live cell apoptosis bioassay (3-6). Loss of apoptosis competence was found in the nonneoplastic colonic mucosa of ∼50% of patients with colon cancer and this loss was patchy in its distribution. A loss in apoptosis competence often results in increased genomic instability (7) and can contribute to carcinogenesis (8). The live cell bioassay is, however, time consuming and cumbersome and requires obtaining fresh biopsies and the use of a CO₂ incubator. In addition, an experienced microscopist is necessary to evaluate the morphologic features of apoptotic cells. A more practical biomarker that would correlate with apoptosis resistance and could be assessed in routine formalin-fixed, paraffin-embedded biopsies is, therefore, clearly needed. The development of a biomarker that is based on the molecular and cellular pathologic events associated with the early stages of colon carcinogenesis would be desirable, because it would be more amenable to meaningful interpretation than a biomarker whose relationship to carcinogenesis is unknown or unclear.

The mitochondrion is a major organelle responsible for the downstream execution phase of apoptosis (9-17). Cytochrome c oxidase, in particular, is the rate-limiting step of the mitochondrial electron transport chain (18) and represents a molecular switch that induces apoptosis under energy stress conditions (19). Of the 13 subunits that comprise cytochrome c oxidase, subunit I is responsible for the control of apoptosis through phosphorylation/dephosphorylation events (20). We have recently found that the inhibition of mitochondrial electron transport at complex IV (cytochrome c oxidase) by sodium azide in vitro increases resistance to sodium deoxycholate–induced apoptosis.⁷

In the present study, we compared the level of expression of cytochrome c oxidase subunit I in the nonneoplastic colonic mucosa of colon cancer patients with the colonic mucosa of low-risk individuals using immunohistochemistry on formalin-fixed, paraffin-embedded tissues. We determined that cytochrome c oxidase subunit I protein expression was markedly decreased in the nonneoplastic colonic mucosa of most of the colon cancer patients and seems a potentially important hypothesis-driven biomarker of colon cancer risk. In addition, there was a high incidence of a striking crypt-restricted loss of cytochrome c oxidase subunit I, an indication of stem cell mutation, in some biopsy sites from patients with colon cancer compared with low-risk subjects. This indicates that high mutational load in mitochondrial DNA (mtDNA) in the “normal” colonic mucosa may be a major risk factor in colon carcinogenesis. Increased variability in the extent of crypt-restricted loss of cytochrome c oxidase subunit I expression was also a statistically significant feature of the neoplasia group. The importance of these findings to biomarker development prompts this report.

All tissues were obtained with informed consent, using a form approved by the University of Arizona Institutional Review Board. Colonic mucosal samples were fixed in 10% buffered formalin and paraffin-embedded for immunohistochemistry. Samples that were stressed ex vivo for assessment of apoptosis competence were prepared as previously described (5, 6).

The level and pattern of expression of cytochrome c oxidase subunit I in nonneoplastic colonic crypts was assessed using standard immunohistochemical methods, as previously described (21-24). Formalin-fixed and paraffin-embedded tissues were cut into 4-μm sections, deparaffinized, and rehydrated. Antigen retrieval was done by microwave exposure in 0.1 mol/L citrate buffer (pH 6.1). Endogenous peroxidase activity was blocked by incubation in 1% hydrogen peroxide in methanol for 30 minutes, and the sections were then rinsed with distilled water and PBS. To prevent nonspecific binding, the slides were incubated with normal rabbit serum (1.5%) for 60 minutes. The primary mouse monoclonal antibody against cytochrome c oxidase subunit I was obtained from Molecular Probes, Inc. (Eugene, OR) and added at a dilution of 1:250 in 2% bovine serum albumin/PBS for 1 hour. After rinsing with PBS, the slides were incubated with biotinylated rabbit anti-mouse secondary antibody IgG F(ab)₂ (DAKO Corp., Carpinteria, CA) for 30 minutes at a dilution of 1:400 in 2% bovine serum albumin/PBS. Immunocontrol slides were prepared by replacing the primary antibody with mouse IgG_2a at the same protein concentration as the primary antibody. After rinsing in PBS, Vectastain Elite avidin-biotin complex method kit (Vector, Burlingame, CA) was used according to the manufacturer's instructions. Color was developed by applying diaminobenzidine tetrachloride (Sigma, St. Louis, MO) supplemented with 0.04% hydrogen peroxide. Sections were counterstained with hematoxylin (Sigma), dehydrated in a graded series of ethanols followed by xylene and mounted using Cytoseal (VWR Scientific, West Chester, PA).

All well-oriented crypts from multiple sites of the nonneoplastic colonic mucosa of patients in different risk groups for colon cancer were evaluated for crypt-restricted loss, segmented loss of immunostaining, or overall decreased immunostaining of cytochrome c oxidase subunit I. Whole crypts that showed most of the longitudinal section of the crypt starting at the lumenal surface and extending to the base of the crypt were scored for crypt-restricted loss of immunostaining. An example of crypt-restricted loss of cytochrome c oxidase subunit I immunostaining (from a patient with colon cancer) is shown in Fig. 1A. The arrows indicate the sharp demarcation between the crypt showing complete loss of cytochrome c oxidase subunit I expression and adjacent crypts showing positive cytochrome c oxidase subunit I immunostaining. Segmented loss of immunostaining is one in which there is strong immunostaining of cells along one or both sides of the crypt adjacent to areas where there is complete absence of staining (Fig. 1B). The strong immunoreactivity using this monoclonal antibody makes the scoring very easy. The percent of crypts with aberrant immunostaining pattern showing crypt-restricted loss, segmented loss of immunostaining, and overall decreased immunostaining was then calculated.

Freshly obtained colon mucosal samples (see Fig. 2) were incubated at 37°C for 3 hours in a CO₂ incubator in the presence of 1 mmol/L deoxycholate, as previously described (5, 6). The tissue was then fixed in half-strength Karnovsky's fixative, post-osmicated, dehydrated, and embedded in epoxy resin. One-micrometer sections were prepared and stained with a polychrome stain (4-6). At least 200 goblet cells were scored for the presence or absence of apoptotic morphology, as previously described (5, 6).

Formalin-fixed, paraffin-embedded tissue was cut into 4-μm sections. The sections were deparaffinized and rinsed in 100% ethanol, 70% ethanol, and PBS. The tissue was stained with nonyl acridine orange according to the method of Cossarizza et al. (25). The sections were incubated for 10 minutes at room temperature with 10 μmol/L nonyl acridine orange (Molecular Probes) before washing twice with PBS. A coverslip was mounted with Fluorescent Mounting Medium (DAKO) before capturing images with a Nikon confocal microscope. Because nonyl acridine orange assesses cardiolipin content of mitochondria (26, 27), relative fluorescence levels are an indication of mitochondrial mass.

This is a pilot study to explore relationships of aberrancies in cytochrome c oxidase subunit I expression, apoptotic index, and colon cancer risk. Data were available on 46 samples from a total of 16 individuals. Given the likely correlation across samples within an individual, and the variable number of samples from each individual, we used mean values for each individual for this exploratory analysis (i.e., the unit of analysis was the individual patient). Values of measurements were compared between neoplasia patients and nonneoplasia patients using t tests with unequal variances. Significance testing was used only to explore the data; Ps should be interpreted cautiously, given both the limited statistical power in such a pilot study plus the multiple comparisons done. The relationships of aberrancies in cytochrome c oxidase subunit I expression and apoptotic index were explored using linear regression and Pearson product-moment correlation coefficients. Statistical analyses were done using JMP software (Macintosh version 5) from SAS Institute, Inc. (Cary, NC).

We evaluated the immunostaining pattern of cytochrome c oxidase subunit I in the nonneoplastic colonic mucosa of 45 tissue specimens obtained from surgically resected colonic segments and one tissue specimen obtained during a colonoscopy procedure. The tissues were obtained from 16 patients, with or without colonic neoplasia. Three different aberrant patterns of cytochrome c oxidase subunit I immunostaining were found: (a) crypt-restricted loss (Fig. 1A and Fig. 3C-D), (b) segmented loss of immunostaining (Fig. 1B), and (c) overall decreased immunostaining (Fig. 3C-D).

Figure 3A and B are representative images of crypts from a control patient without colonic neoplasia showing normal levels of cytochrome c oxidase subunit I immunostaining. In Fig. 3A, note the intense positive stain for cytochrome c oxidase subunit I throughout the length of the crypt with even heavier staining at the surface epithelium and upper portion of the crypt. Figure 3C and D are representative images of crypts from the nonneoplastic colonic mucosa of a patient with colon cancer. A crypt with total absence of staining (crypt-restricted loss; Fig. 3C, arrows) is clearly demarcated from the stained crypts to its left and right. This crypt-restricted loss staining pattern, also shown at higher magnification in Fig. 1A, is likely a reflection of stem cell mutation. Segmented loss of immunostaining (Fig. 1B) is where there is strong immunostaining of cells along portions of one or two sides of the crypt adjacent to areas with complete absence of staining. The third aberrant pattern, overall decreased immunostaining, is shown in Fig. 3C (crypts to the left and right of arrows) and Fig. 3D (middle and left crypts). The overall decreased immunostaining is particularly evident at the bases of the crypts of the patient with cancer shown in Fig. 3D when compared with the staining at the bases of the crypts from a patient with no evidence of colonic disease shown in Fig. 3B.

The cytochrome c oxidase subunit I immunostaining results (% crypt-restricted loss, % segmented loss of immunostaining, and % overall decreased immunostaining) for the 46 samples of normal mucosa from five control subjects without neoplasia (three normal and two diverticulitis) are shown in Table 1, and those for the nonneoplastic colonic mucosa from 11 patients with neoplasia [three patients with large tubulovillous adenomas (at elevated risk for cancer) and eight patients with colonic adenocarcinoma] are shown in Table 2. Multiple samples were obtained from each patient to determine variability within a patient sample. For subsequent analyses of apoptotic index, the minimum of the two measurements for each sample was used as the value of apoptotic index for that sample. [Note: performing calculations using the minimum value (i.e., the one with potentially the greater decrease in competence) allows “patchiness” of the field defect to be taken into consideration (6)]. One of the nonneoplasia individuals had only a single sample, and apoptotic index was not measured for that sample; thus, this individual did not contribute either to analyses of apoptotic index nor to analyses of within-person heterogeneity.

Analysis of aberrancies in expression of cytochrome c oxidase subunit I (Table 3) showed that, compared with control subjects, the patients with neoplasia had a significantly higher incidence of crypts having overall decreased immuno-staining of cytochrome c oxidase subunit I (1.7 versus 22.8, P = 0.03) and a higher incidence of crypt-restricted loss (0.6 versus 3.2, P = 0.06). The incidence of segmented loss of immunostaining of cytochrome c oxidase subunit I expression was low and was similar in the two groups. Combining these results, the incidence of normal expression of cytochrome c oxidase subunit I (i.e., with none of the three types of abnormality) was 96.7 ± 1.2% in control subjects versus only 73.2 ± 8.7 % in patients with neoplasia (P = 0.02). It should be noted that a defect in cytochrome c oxidase subunit I immunostaining was not detected in all biopsy sites from the patients with neoplasia, indicating a “patchiness” in the cytochrome c oxidase subunit I field defect. The aberrant staining was not a reflection of being close to the tumor, because crypt-restricted loss of cytochrome c oxidase subunit I immunostaining was seen as far as 15 cm away from one of the tumors, and 8 and 9 cm away from tumors in other patients.

For control subjects, the mean apoptotic index was 51.5% compared with only 38.2% for patients with neoplasia (P = 0.03; Table 3), reflecting decreased apoptosis competence in the nonneoplastic colonic mucosa of these patients. We also investigated whether there was more within-person heterogeneity (i.e., across the samples within individuals) for patients with neoplasia compared with control subjects. The results (Table 3) showed that the neoplasia cases had greater across-sample SD (i.e., increased variability) for % crypt-restricted loss (P = 0.02) and for apoptotic index (not statistically significant; P = 0.15) when compared with those of the nonneoplasia cases. However, the SD of % crypt-restricted loss tended to increase with increasing mean % crypt-restricted loss (although not significant by linear regression, P = 0.19) and the SD of apoptotic index increased with decreasing mean apoptotic index (P = 0.04, linear regression). When the SD was divided by the mean value (for % crypt-restricted loss) or by 100 minus the mean value (for apoptotic index), the differences in sample variability (as a fraction of the mean) between the neoplasia and nonneoplasia groups of patients were small and not statistically significant (Table 3). Nevertheless, the extent of variability among patients with neoplasia is informative; the increasing variability (across samples) with increasing mean abnormality is consistent with a patchy defect.

The relationship of apoptotic index to % of crypts with normal cytochrome c oxidase subunit I immunostaining, using the mean for each patient, is shown in Fig. 4. Linear regression gave a statistically significant association (correlation = 0.69; P < 0.01). As an exercise in hypothesis generation from these pilot data, note that the four control subjects all had a mean incidence of normal crypts of >92% (Fig. 4), with no individual value below 88% (Table 1). However, 7 of the 11 (64%) neoplasia patients had mean % normal crypts of <88, and for these seven patients, no individual sample had a value of >88 (Table 2). One additional neoplasia patient had values (% crypts with normal cytochrome c oxidase subunit I immunostaining) of 89.8 and 88.6 for the two samples. The four control subjects all had a mean apoptotic index of >42 (Fig. 4), with no individual value below 38 (Table 1). For this measure, 5 of the 11 (45%) neoplasia patients had a mean apoptotic index of <38, but individual samples for many of these patients did have values exceeding 38. Overall, these results suggest that aberrant cytochrome c oxidase subunit I expression may be a better biomarker than loss of apoptosis competence. Because this analysis used data-derived cutoffs from a small sample, further investigation and separate validation in a larger study are required.

Because cytochrome c oxidase subunit I is a mitochondrial protein, the loss of cytochrome c oxidase subunit I expression could have been caused by a marked decrease in the number of mitochondria. Mitochondrial mass has not previously been evaluated in situ as cells differentiate along the crypt length. Therefore, we evaluated the nonneoplastic mucosa of an index patient with colon cancer who exhibited a marked overall decreased immunostaining and high frequency of crypt-restricted loss of cytochrome c oxidase subunit I to test this hypothesis. Additional paraffin-embedded 4-μm sections of the colonic epithelium from the normal and colon cancer patients whose crypts are shown in Fig. 3, were stained with 10 μmol/L nonyl acridine orange (Fig. 5). Nonyl acridine orange is a fluorescent stain specific to mitochondria, because it binds to cardiolipin (26, 27), a lipid found only in mitochondrial membranes. Nonyl acridine orange staining does not depend upon the mitochondrial membrane potential; therefore, differences in staining reflect alterations in mitochondrial mass not mitochondrial function. In Fig. 5, positive fluorescence of nonyl acridine orange is pseudocolored green; however, very intense nonyl acridine orange staining is indicated by the white areas. In the normal patient, there is an increase in mitochondrial mass at the surface epithelium and upper part of the crypt (white areas) compared with the base (Fig. 5A-B). It can be seen in Fig. 5 that there is no loss of mitochondrial mass in the nonneoplastic colonic mucosa of a patient with colon cancer (Fig. 5C-D). Instead, there is a definite increase in mitochondrial mass in the nonneoplastic colonic mucosa of the patient with cancer, as evidenced by the more extensive white areas throughout the entire crypt. The loss of cytochrome c oxidase subunit I immunostaining in the colonic crypts from patients with colon cancer is, therefore, probably caused by a reduction in gene expression rather than a loss of mitochondria.

Colon cancer is the second leading cause of cancer deaths around the world. Because the cure rate for metastatic colon cancer is <50%, early diagnosis and the identification of patients at high risk for colon cancer is of paramount importance. The identification of biomarkers that are both specific and sensitive is crucially needed to target individuals for either more frequent colonoscopies, recommendations for lifestyle changes, and/or chemopreventive strategies. The data presented in this study indicate that we may have identified such a biomarker. We found a generalized loss of cytochrome c oxidase subunit I expression in many of the nonneoplastic colonic crypts of patients with colon cancer. In addition, we observed dramatic crypt-restricted loss of cytochrome c oxidase subunit I immunostaining, which has not been previously described in association with the presence of colonic neoplasia.

Because mitochondria are major mediators of apoptosis (9-17), our finding of defects in the expression of a mitochondrial enzyme known to be associated with the modulation of apoptosis (19) in the nonneoplastic colonic mucosa of patients with colon cancer, is likely to be very significant. The present study, indicating a statistically significant correlation between decreased cytochrome c oxidase subunit I expression and apoptosis resistance in the nonneoplastic mucosa of patients with colon cancer, provides support for a possible cause-and-effect relationship between these two biological processes.

It has been shown that crypt-restricted expression of proteins (28, 29), loss of expression of proteins within crypts (30, 31), or loss of histochemical activity within crypts (32-34) reflect stem cell mutations. Several animal models have been described (see review by Garcia et al.; ref. 35) in which stem cell behavior was followed over time after mice were given a single dose of mutagen. Partial, or segmented, mutant phenotypes appear first. Later, there is an increase in the number of crypts showing a completely or wholly mutated phenotype (i.e., crypt-restricted loss in expression of the affected protein). The crypt-restricted pattern of expression is, therefore, the consequence of an earlier mutational event (28, 36, 37). However, in humans, as distinct from animal models, documentation of crypt-restricted loss of any protein is sparse (35). In particular, crypt-restricted loss of cytochrome c oxidase subunit I, as described in this study by assessing the entire longitudinal length of the crypt, has not been previously reported.

The only previously documented crypt-restricted loss of protein expression in human colonic crypts is the complete loss of sialyl O-acetyltransferase from individual crypts (30, 31). Because of the crypt-restricted loss of ability to acylate O-sialomucin, the goblet cells in an entire single crypt stain positively with mild periodic acid Schiff (38), or negatively by the periodate-borohydride/potassium hydroxide saponification/periodic acid Schiff method which stains O-acetyl sialic acid residues (39). This crypt-restricted staining with mild periodic acid Schiff was observed after X-irradiation therapy for rectal cancer (31) or in the nonneoplastic colonic mucosa of patients with both right- or left-sided sporadic colorectal carcinomas (38). Another type of mutation-related alteration in protein expression is an increase in crypt-restricted metallothionein protein immunostaining (28, 29, 40). In these reports, the frequency of crypt-restricted expression of increased metallothionein or reduced mild periodic acid Schiff staining was much lower than what we found in our crypt-restricted loss of cytochrome c oxidase subunit I immunostaining. We found an average of 2.8% crypt-restricted loss of cytochrome c oxidase subunit I immunostaining in the crypts of the 11 patients with colonic neoplasia and up to 13.7% crypt-restricted loss at one biopsy site (see Table 2). In contrast, Campbell et al. (38) reported a maximum of 0.0044% of the crypts showing wholly involved (i.e., crypt restricted) loss of O-acetyltransferase in the nonneoplastic colonic mucosa of patients with left-sided colorectal cancer after examining ∼10,000 crypts. The increased frequency of crypt-restricted metallothionein expression in the nonneoplastic colonic mucosa at the resection margins of patients with colonic adenocarcinomas was reported as being “infrequent” (40). In the animal study of crypt-restricted increase in metallothionein expression, it was reported that after a single dose of the mutagen dimethylhydrazine, the frequency of crypt-restricted metallothionein expression was <0.00129% (28). Thus, the crypt-restricted loss of cytochrome c oxidase subunit I in the present study seems 640- to 2,200-fold more frequent than the crypt-restricted loss of O-acetyltransferase and the crypt-restricted increase in metallothionein expression. (Note: the metallothionein estimate is based only on animal models using a mutagen). The differences in the frequency of apparent stem cell mutations reported in the literature (O-acetyltransferase and metallothionein) compared with our data on cytochrome c oxidase subunit I, may reflect the fact that O-acetyltransferase and metallothionein are nuclear-encoded genes, whereas cytochrome c oxidase subunit I is encoded by the mitochondrial genome. The increased frequency of crypts showing crypt-restricted loss of cytochrome c oxidase subunit I expression may reflect greater production of reactive oxygen species within mitochondria during colon carcinogenesis, which would preferentially induce mutations in mtDNA. In contrast to the loss of the differentiation marker, O-acetyltransferase, loss of cytochrome c oxidase expression may be related to the process of apoptosis resistance. Apoptosis resistance may result in greater mutagenesis (7) and thus is likely to promote carcinogenesis.

We have found in the present study (using an index case) that the loss of cytochrome c oxidase subunit I is not caused by a loss of mitochondria, using the cardiolipin dye, nonyl acridine orange. In fact, mitochondrial mass was increased in the tissue showing loss of cytochrome c oxidase subunit I expression. This finding is consistent with persistent exposure of colonic epithelial cells to oxidative stress during the early stages of colon carcinogenesis, because oxidative stress is known to increase mitochondrial mass (20, 41).

Crypt-restricted loss of cytochrome c oxidase subunit I probably reflects stem cell mutations, which may be caused by certain dietary factors. Bile acids are known tumor promoters in animal models (42) and are elevated in the feces of colon cancer patients and patients with adenomatous polyps (43). Numerous studies have shown that hydrophobic bile acids can induce oxidative DNA damage (e.g., refs. 44, 45), which implies that bile acids are probably carcinogens in humans (reviewed in ref. 46). Bile acids also damage mitochondria (47, 48), resulting in increased oxidative stress (49) and reduced mitochondrial membrane potential (47, 50, 51).

Because cytochrome c oxidase subunit I is one of the 13 proteins encoded by mtDNA (see complex IV, Fig. 6, modified from refs. 52, 53), we hypothesize that the crypt-restricted loss of mitochondrial-encoded cytochrome c oxidase subunit I in the nonneoplastic colonic mucosa of patients with colon cancer or at high risk for colon cancer may be due to specific mitochondrial mutations. In fact, mutations in all three mitochondrially encoded subunits of cytochrome c oxidase have been identified in human colorectal tumors (54) and most recently in colonic crypts of the nonneoplastic colonic mucosa of aged individuals (34).

It may be initially puzzling as to how a stem cell mutation can account for the observation of crypt-restricted loss of cytochrome c oxidase subunit I, given that cytochrome c oxidase subunit I is a mitochondrial gene, there are several copies of mtDNA per mitochondrion, many mitochondria per cell, and multiple stem cells per crypt. However, we can offer an explanation in terms of natural selection. Central to this explanation are the normal process of mitochondrial turnover (Fig. 7A) and the abnormal turnover of mutant mitochondria and aberrant stem cells (Fig. 7B). The latter involves the generation of multiple copies of mutant mtDNA in a single mitochondrion, the replicative advantage of mutant mitochondria in single cells, the generation of a stem cell with an apoptosis-resistant phenotype, the repopulation of stem cells in the “stem cell niche” (55) with mutant stem cells having a proliferative advantage, and the generation of a crypt showing crypt-restricted loss of cytochrome c oxidase subunit I. A plausible sequence of events is now presented.

1. It has been well documented that effete/damaged mitochondria that generate excessive reactive oxygen species causing lipid peroxidation (56) lose their membrane potential, resulting in the opening of the mitochondrial permeability transition pore (Fig. 7A). This initiates the process of autophagy (57, 58), which is the main degradative process in the cell responsible for the removal of damaged mitochondria and other organelles (for reviews of the autophagic process, see refs. 59-68). It is unclear what the actual autophagic signals are. They may include the release of proautophagic signals through the mitochondrial permeability transition pore (ref. 63; Fig. 7A) and/or the presentation of abnormal aggregates of damaged, misfolded proteins on the outer mitochondrial membrane (58). These abnormal aggregates may then interact with proautophagic proteins (i.e., beclin; refs. 69, 70), which act as molecular recognition elements (63) and elicit the formation of the autophagosome. The autophagosome then fuses with the lysosome, forming the autophagolysosome (Fig. 7A). Damaged mitochondria are digested by the lysosomal acid hydrolases. The autophagic process elicits a compensatory increase in mitochondrial biogenesis to maintain the normal population of mitochondria (71-73).

2. Next, consider a scenario in which the CcOI gene is mutated in a mtDNA molecule within a given mitochondrion. The mutant mtDNA replicates along with wild-type mtDNA and, upon mitochondrial division, copies segregate to daughter mitochondria. After a succession of mitochondrial divisions, the mitochondrion which, by chance, has a higher proportion of the mutant mtDNA will exhibit a reduction in the cytochrome c oxidase subunit I protein (Fig. 7B).

3. A mutational defect in any of the 13 mtDNA-encoded respiratory subunits (see Fig. 6) can lead to respiratory chain dysfunction (72, 74). If the cytochrome c oxidase subunit I mutation and loss of cytochrome c oxidase subunit I protein decreases reactive oxygen species formation by the respiratory chain, there will be less lipid peroxidation, and the permeability transition pore will remain closed. This leads to the failure to elicit autophagy (58), resulting in a replicative advantage to the mutant mitochondria and the accumulation over time of homoplastic mutant mitochondria within the cell (ref. 75; Fig. 7B). It has been shown, through cell fusion experiments and PCR amplification experiments in tissue from patients with mitochondrial disease, that mutant mtDNA can be functionally dominant over wild-type mitochondrial genomes (76), can have a replicative advantage (77), and can be clonally expanded (54, 78).

4. A decrease in the ATP/AMP ratio caused by mitochondria with respiratory chain dysfunction may then increase mitochondrial biogenesis through the activation of AMP kinase and an increase in nuclear and mitochondrial respiratory-related transcription factors (79, 80).

5. Once a cell within the stem cell niche of a crypt has a defect in cytochrome c oxidase subunit I in most of its mitochondria, that stem cell may have reduced ability to undergo apoptosis. A colonic stem cell containing mutant mitochondria that cannot trigger the apoptotic response would confer apoptosis resistance and a proliferative advantage at the cellular level. Anchorage of such an apoptosis-resistant stem cell at the base of the crypt could eventually lead to the repopulation of the stem cell niche with cells having a homoplastic mitochondrial mutuation (i.e., within the CcOI gene; Fig. 7B). The subsequent division of these mutant stem cells would then result in crypt-restricted loss of cytochrome c oxidase subunit I.

In conclusion, the crypt-restricted loss and/or markedly decreased overall immunostaining of cytochrome c oxidase subunit I may prove to be reliable biomarkers of colon cancer risk, because they occur at elevated frequency in the nonneoplastic colonic mucosa of patients with, and at high risk for, cancer. A particular value of reduced cytochrome c oxidase subunit I expression as a biomarker is that it is hypothesis driven, because reduction or loss of cytochrome c oxidase subunit I activity likely contributes to apoptosis resistance, a preneoplastic field defect in colonic epithelia previously identified by our group, using a live cell ex vivo bioassay (3, 5, 6).