Possible link between the mitochondrial redox state alteration and cancer transformation
We imaged the mitochondrial redox indices of the PTEN-null mouse model using the redox scanner to look for potential new biomarkers for pre-malignancy detection. To support their rapid growth and survival in stressful and dynamic microenvironment, cancer cells constitutively activate their key oncogenic signaling pathways which converge to adapt/alter core cellular metabolism of all four major classes of macromolecules: carbohydrates, proteins, lipids, and nucleic acids. As one of the key oncogenic signaling pathways, PI3K/AKT pathway has profound effects on the metabolism of cancer cells
. It regulates glucose uptake and utilization and its activation renders cells dependent on high levels of glucose flux
. The downstream molecule NF-κB in the PI3K/AKT pathway controls the balance between the utilization of glycolysis and mitochondrial respiration
. Studies show that up to 60% of pancreatic cancers have either amplified or activated AKT2 kinase
[45–47]. PTEN is a negative regulator of PI3K/AKT pathway. PTEN mutations amplify PI3K/AKT signaling and promote tumorigenesis. For example, it is shown that genomic deletion of PTEN is associated with prostate tumor metastatic potential
. However, the in vivo metabolic evaluation on the role of activated PI3-K/AKT signaling in the initiation or progression of pancreatic cancer has not been reported.
Our results showed that the PTEN-null pancreases had many regions with higher values of the Fp redox ratio (as typically seen in Figures
3). The heterogeneity markers such as the SDs or the histogram widths (ci) of the Fp redox ratio, Fp/NADH, and NADH/Fp are significantly larger in the PTEN-null pancreases, whereas the tissues in the control group were relatively more homogeneous with significantly smaller SDs and smaller Gaussian curve width ci (Tables
3). Higher level of metabolic heterogeneity is consistent with previous observations on various pre-cancer tissues
[37, 38]. Therefore, it is likely that heterogeneous mitochondrial redox state is associated with cancer transformation. This indicates that (pre)malignant tissues may exhibit significant mitochondrial redox abnormality in localized spots before fully transformed.
We also observed lower levels of Fp and NADH in the PTEN-null group (Table
2). Decreased Fp and NADH is probably partially due to decreased cell density in premalignant tissues which had various sizes of ductal structures and cystic dilations
. The ratio indices that represent the metabolic and redox states of the tissues, Fp/NADH, NADH/Fp, or Fp/(Fp + NADH) and their standard deviations are less sensitive to the variability of cell density. A few other factors could affect Fp and NADH fluorescence signals, such as blood circulation and instrument fluctuations. Ratiometric indices such as the redox ratios are largely insensitive to these factors. Notably, only the SDs of the ratio indices distinguished between two groups whereas the SDs of Fp and NADH did not. This suggests that the heterogeneity of the ratio indices may be a better marker than the heterogeneity of Fp and NADH for the identification of premalignant tissues.
The result from this study also indicates that PTEN deletion shifted the tissue to a more oxidized state. Significantly higher Fp redox ratio and Fp/NADH were found in the PTEN-null group compared to the control group (Table
2). Higher Fp redox ratio or Fp/NADH was not reported previously by other researchers on various pre-cancer tissues
[36–38]. For example, one of the studies imaged the redox ratios at various tissue depths for precancerous epithelia of oral cancer
, and no significant difference was found in the redox ratios when comparing global averages between the precancerous and normal tissues. This is exactly what we have seen using Method I analysis. However, we saw statistically significant difference when using Method II that takes tissue depth as a covariate in GLM analysis. Another study
 measured the 3D distribution of the redox state in fast-frozen human cervical tissues that typically consisted of an upper epithelium and a lower stroma. When compared to the normal tissue, the dysplastic epithelia had decreased redox ratio whereas the stroma appeared to have increased redox ratio (estimated based on their reported data
). However, it was not reported whether there was any difference in the overall averages of the redox ratios of the entire tissue block. The other study
 measured the autofluorescence signals to investigate the redox state in the clinical cervical tissue samples that have been kept under temperatures above 0°C for 1.5-5 hours before imaging. Reduced redox ratios were found in the dysplastic epithelia of one third of the patient samples compared to the paired normal samples. During the long waiting period before imaging, the intrinsic fluorescence signal and the redox state in the tissue samples might have changed significantly and differential changes between the precancerous and the normal tissues might have occurred. In contrast, the mice’s pancreases used in our study were resected and dropped in liquid nitrogen within two seconds, and have been kept in liquid nitrogen for storage and during the redox scanning process. Their mitochondrial redox states should be the same as or very similar to the in vivo condition due to the snap-freezing procedure.
Nevertheless, the slightly higher Fp redox ratio in the PTEN-null group appears to be consistent with our previous studies on the redox states of the normal tissue and indolent and aggressive tumor tissues
[26, 32–34]. The normal muscle tissues had lower Fp redox ratio compared with the cancer tissues. The aggressive tumors exhibited a readily-recognized bi-modal distribution of their redox indices (e.g. Fp, NADH, Fp redox ratio) with substantial regions having significantly higher Fp redox ratio, i.e., more oxidized redox state, whereas the indolent tumors had relatively homogeneous and reduced redox states. The more aggressive the tumors were, the more heterogeneous they were and higher Fp redox ratio in the oxidized regions. Thus, the results of this study provide more evidence to a link between malignant transformation/progression and a more oxidized and heterogeneous mitochondrial redox state.
However, the role of the mitochondrial redox state in cancer transformation/progression remains largely unclear. Connections between the redox potentials (or NADH levels) and malignancy have been implicated in some studies
[49–54]. High levels of free radicals/oxidants are known to generate oxidative damages and genetic mutations that may drive cancer transformation/progression. The mitochondrion is one of the major sources of free radical generation. Mitochondrial dysfunctions have been shown to induce overproduction of reactive oxygen species (ROS), which increases metastatic potential of cancer cells
[51, 55] and drives tumor transformation/progression in surrounding tissues in conjunction with oncogenic Ras
. However, ROS-independent pathways may also exist
. Nevertheless, high levels of free radicals/oxidants do not necessarily indicate oxidized redox potential. The high levels of oxidants in tumors are often counter-balanced by high levels of reductants such as vitamin C, reduced glutathione, and reduced nicotinamide adenine dinucleotide phosphate (NADPH)
[53, 58–60]. Cellular redox potential, i.e., the balance between reductants and oxidants, is mediated by multiple intracellular redox couples, e.g. NAD+/NADH, NADP+/NADPH, and oxidized/reduced glutathione. Redox biology is further complicated by the uneven distributions of redox systems in subcellular compartments (cytosolic, nuclear, mitochondrial, etc.). The mitochondrial redox potential may or may not be coupled with the cytosolic redox potential
. If we can accurately and precisely quantify the intracellular redox potential in each subcellular compartment, then it is possible to have a better understanding of the connections between mitochondrial redox state and cancer transformation/progression as evidenced by our studies.
Importance of imaging intratumor heterogeneity by submillimeter imaging methods
It is known that concentrations of crucial nutrients such as glucose, glutamine, and oxygen are spatially and temporally heterogeneous in solid tumors
[9, 10]. Intratumor heterogeneity is an important factor for cancer diagnosis and treatment. The variability among subpopulations of tumor cells is well known for tumor progression and such heterogeneity is associated with malignancy and metastasis
[62–65] as well as treatment failure
. Studies also showed that tissue heterogeneity is indicative of malignant phenotypes
[67, 68]. Imaging tumor heterogeneity may grade tumor invasive/metastatic potential and yield important prognostic and/or predictive biomarkers
[32, 33, 69]. Promisingly, imaging pre-cancer tissues also revealed the existence of metabolic heterogeneity in dysplastic/premalignant tissues
Since the tissue metabolic heterogeneity may be on a submillimeter scale
, other metabolic biomarkers such as FDG-PET with a spatial resolution on the order of millimeters
[70, 71] are incapable of resolving it. Therefore, it is important to study tissues by using a high resolution imager. The redox scanner
[24, 27] we employed has a spatial resolution down to 50 × 50 × 20 μm3. As shown by the present study, the submillimeter high resolution redox imaging of tissue heterogeneity enhances the capability to identify biological factors that are important for cancer transformation but may not be seen by low resolution imaging techniques or the measurements of tissue homogenates. It should also be noted that the snap-freezing of the pancreases right after resection ensured the redox state of the frozen tissues is the same as or very similar to that of the in vivo condition.
Selection of proper analysis methods to account for tissue heterogeneity
In our previous studies, when averaging over entire tumor, the heterogeneity information was largely lost, resulting in less significant or insignificant difference in Fp redox ratio between the aggressive and indolent tumors
[32, 33]. Consistently, we did not see significant difference in the Fp redox ratio between the premalignant and the control group after the global averaging for each pancreas using Method I that largely ignores the spatial heterogeneity of these redox indices. However, the redox images show that tissue heterogeneity exists in three dimensions in both the PTEN-null and control groups, with the PTEN-null group having a distinctly higher level of heterogeneity as evidenced by both image visualization (Figures
3) and the quantifications of the standard deviations and widths of the Gaussian fits. The global averaging in Method I effectively removed the information on tissue heterogeneity and also reduced the number of independent data for statistical analysis. Thus, global average does not truly represent the heterogeneous nature of the diseased tissues, which may be critically important for distinguishing between groups.
Compared to Method I, Method II recognizes tissue heterogeneity in one of the three dimensions by using tissue depth (z-direction) as a covariate for the statistical analysis. Each tissue section has a tissue depth that is the distance between the top section and the specified section. Increased statistical significance was observed using Method II, yielding much smaller p values for the differences of the means between the PTEN-null and control groups. For example, a significant p value of 0.01 was reached for comparing the means of the Fp redox ratios from Method II whereas Method I failed to statistically differentiate Fp redox ratios between two groups. The difference between Method I and II shows the importance of selecting proper statistical analysis method for comparing multi-slice imaging results.
This result also shows the advantage of imaging measurements compared with the common practice that measures tissue homogenates where tissue has been ground and mixed to be homogenized
[61, 72, 73]. We observed high heterogeneity of the redox state in both coronal and transversal planes of PTEN-null pancreases and along tissue depth. As heterogeneity is one of the very important characteristics of pathological tissues including precancerous tissues, the more tissue heterogeneity information is acquired, the better they are characterized. Suitable statistical analysis methods are also needed to account for the heterogeneity in three dimensions.
One of the shortcomings of the study is that the spatial resolution of the redox images of 100 μm used in this study is still not capable of fully resolving the pancreas premalignant lesions involving the highly proliferative ductal formations (size on the order of ~200 μm). In future, we may use CCD based redox imaging techniques to advance the study at a higher spatial resolution. We may also co-register the redox images with the histological staining and incorporate the in-depth studies on the biology to identify the sources of those more oxidized spots.