Bioimaging of Fluorescence-Labeled Mitochondria in Subcutaneously Grafted Murine Melanoma Cells by the “In Vivo Cryotechnique”

Abstract

The microenvironments of organs with blood flow affect the metabolic profiles of cancer cells, which are influenced by mitochondrial functions. However, histopathological analyses of these aspects have been hampered by technical artifacts of conventional fixation and dehydration, including ischemia/anoxia. The purpose of this study was to combine the in vivo cryotechnique (IVCT) with fluorescent protein expression, and examine fluorescently labeled mitochondria in grafted melanoma tumors. The intensity of fluorescent proteins was maintained well in cultured B16-BL6 cells after cryotechniques followed by freeze-substitution (FS). In the subcutaneous tumors of mitochondria-targeted DsRed2 (mitoDsRed)-expressing cells, a higher number of cancer cells were found surrounding the widely opened blood vessels that contained numerous erythrocytes. Such blood vessels were immunostained positively for immunoglobulin M and ensheathed by basement membranes. MitoDsRed fluorescence was detected in scattering melanoma cells using the IVCT-FS method, and the total mitoDsRed volume in individual cancer cells was significantly decreased with the expression of markers of hypoxia. MitoDsRed was frequently distributed throughout the cytoplasm and in processes extending along basement membranes. IVCT combined with fluorescent protein expression is a useful tool to examine the behavior of fluorescently labeled cells and organelles. We propose that the mitochondrial volume is dynamically regulated in the hypoxic microenvironment and that mitochondrial distribution is modulated by cancer cell interactions with basement membranes.

Introduction

The metabolic activity of tumor cells has marked effects on their aberrant proliferation and cell migration, their capacity to evade growth-inhibitory signals, and in the ability to disseminate into distant tissues of living organs (Schulze and Harris 2012). Such metabolic profiles of tumor cells depend on the tissue microenvironment, such as alterations to the local blood supply, which is considered to be spatially and temporarily dynamic (Junttila and de Sauvage 2013). Therefore, living tumor cells are highly affected by the mitochondrial functions that control energy production, intermediate metabolism, and necrotic or apoptotic cell death (Wallace 2005). Mitochondria in animal cells are also a primary source of reactive oxygen species, which can facilitate the development of tumors in living animal organs (Egeblad et al. 2010; Ladiges et al. 2010). On the other hand, mitochondrial functions and energy homeostasis depend on dynamics that regulate the distribution and morphology of mitochondria within cells (Chan 2006; Nunnari and Suomalainen 2012). In addition, mitochondrial morphology varies highly in tumor tissues, and the tumor microenvironment, such as hypoxic situations, can easily influence mitochondrial functions (Arismendi-Morillo 2009; Semenza 2009). However, the common morphological features of mitochondria in the complex histological architecture of tumor tissues are often affected by technical artifacts associated with ischemia/anoxia and tissue shrinkage during conventional preparation steps (Hayat 1989).

The “in vivo cryotechnique” (IVCT) is a powerful tool that has been used to clarify the morphological states of functioning animal organs under blood circulation. It can also prevent technical artifacts caused by tissue-resection and immersion/perfusion fixation, and reveal the in situ morphology of living animal organs (Ohno et al. 2007; Ohno et al. 1996). In addition, IVCT can preserve the rapid alterations to metabolic markers that are vulnerable to ischemia/hypoxia stresses, and also retain the in situ distributions of intracellular/extracellular molecules and extrinsic probes (Huang et al. 2013; Terada et al. 2005; Terada et al. 2009; Terada et al. 2010; Zea-Aragon et al. 2004). Indeed, the morphofunctional correlations between the distribution of functional blood vessels and the expression of hypoxia-induced vascular endothelial growth factor were able to be clearly detected with markedly less tissue shrinkage and no blood vessel collapse when IVCT was followed by freeze-substitution (FS) in recent experiments using xenografted tumor tissues (Bai et al. 2009; Ohno et al. 2008). Since genetically prepared fluorescent proteins have been widely used as molecular tags to reveal the morphofunctional dynamics of cells and organelles as well as the localization and interactions of target proteins (Giepmans et al. 2006), the combination of IVCT with such fluorescent proteins may be useful in understanding organellar dynamics within the functional cells of living animal organs in the context of tissue microenvironments.

The purpose of this study was to use mitochondria-targeted fluorescent proteins in IVCT-prepared specimens and directly visualize mitochondrial distributions within individual melanoma cells relative to the functional blood vessels of tumor tissues. The results showed that the fluorescence signals of multiple fluorescent proteins, including mitochondria-targeted DsRed2 (mitoDsRed), were well maintained after quick-freezing followed by FS, when paraformaldehyde was used as a fixative for FS. Furthermore, the mitochondrial volumes of viable melanoma cells surrounding functional blood vessels were decreased with the expression of hypoxia markers. These mitochondria were often localized in the tumor cytoplasm and processes, which were attached to and extending along basement membranes. These findings also demonstrated that fluorescent protein expression is useful for the histological examination of cell organelles with cryotechniques, and are consistent with the concept that the volume of mitochondria is dynamically regulated in the hypoxic microenvironment and that the distribution of mitochondria is modulated in response to tumor cell interactions with basement membranes.

Materials & Methods

Cell Culture, Plasmid Transfection and Drug Resistance Selection

Mouse B16-BL6 melanoma cells were kindly provided by Dr. I. J. Fidler (Anderson Cancer Center, Houston, TX). The cells were maintained as monolayer cultures in Eagle’s MEM (EMEM; Sigma-Aldrich Japan, Tokyo, Japan) supplemented with 10% fetal bovine serum (FBS). Cultures were kept in a humidified atmosphere of 5% CO₂/95% air at 37C. Cells were transfected with the plasmids after reaching 70–80% confluence.

Melanoma tumor cells cultured on PLL (poly-L-lysine)-treated coverglasses were transfected with pLenti6/V5 vectors (Invitrogen, Carlsbad, CA) expressing the blasticidin-resistance gene and mitoDsRed2 (Ohno et al. 2011), mitoDendra2 (Wang et al. 2008), EGFP, or mCherry (Clontech, Palo Alto, CA) using Lipofectamine 2000 (Invitrogen). Some melanoma cells transfected with pLenti-mitoDsRed were treated with blasticidin (Invitrogen) at a final concentration of 20 µg/ml for 7 days for drug resistance selection.

Quick-Freezing or Immersion Fixation of Cultured Cells

This protocol is outlined in Figure 1A. Cultured and transfected cells for immersion fixation were incubated in either (i) 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) or (ii) 100% ethanol for 6 hr, washed with phosphate-buffered saline (PBS), and embedded in anti-fading reagent containing 70% glycerol, 60 mM phosphate buffer, 10 μM p-phenylenediamine (pH 8.5), and 500 ng/ml DAPI. For quick-freezing followed by freeze-substitution (FS), coverglasses with cultured and transfected melanoma cells were quickly plunged into the liquid isopentane-propane (IP) cryogen (-193C) and then maintained in liquid nitrogen. The coverglasses were transferred into vials containing FS solutions, and these vials were kept in acetone with dry ice (-80C) for 1 hr. Three different FS solutions were used: (iii) dehydrated acetone alone, (iv) acetone containing 2% paraformaldehyde, and (v) acetone containing 2% paraformaldehyde and 0.5% glutaraldehyde. The temperature was gradually increased by incubating the vials at -30C, -10C, 4C, and room temperature for 1 hr each. Samples were washed in PBS and embedded in Vectashield (Vector Laboratories, Burlingame, CA). All samples were observed under the same conditions with a light microscope (BX-61; Olympus, Tokyo, Japan) equipped with a CCD camera. To compare fluorescence intensities, images were taken under the same conditions with a 20× objective lens (N.A. 0.7) and an exposure time of one eighth of a sec. Fading was not observed during the 10–20 sec of imaging under these observation conditions. Images were taken with a 40× objective lens (N.A. 0.85) for detailed observation of individual cells. Two replicates were examined to obtain representative images.

Figure 1.

Experimental design of this study. (A) To examine fluorescence retention after quick-freezing and freeze-substitution (FS), cultured B16-BL6 melanoma cells are transfected to express fluorescent proteins of interest. The cells are then treated by immersion fixation with paraformaldehyde (i, PFA) or ethanol (ii, EtOH), or quickly frozen by plunging cells into isopentane-propane cryogen (-193C) (QF). These quickly frozen cells are fixed by freeze-substitution (FS) at -80C in dehydrated acetone (iii, FS-Ace), acetone containing 2% paraformaldehyde (iv, FS-PFA), or acetone containing 2% paraformaldehyde and 0.5% glutaraldehyde (v, FS-GA). All tumor cells are embedded in anti-fading reagent and observed under a fluorescence microscope. (B) To examine tumor tissues in vivo, melanoma cells transfected with mitochondria-targeted DsRed2 (i) are selected according to antibiotic-resistance (ii) and subcutaneously injected into C57BL/6 mice (iii). After they formed masses (iv, arrowhead), tumor tissues are obtained (v), after the mice are treated with either perfusion fixation followed by alcohol dehydration (PF-DH) or the “in vivo cryotechnique” followed by freeze-substitution (IVCT-FS).

Mouse Model of Melanoma Tumor and Procedures for Tissue Preparation

Five, male, 7-week-old C57BL/6J mice, weighing 20–28 g, were used in the present study. All experimental procedures were in accordance with the Guidelines for Care and Use of Experimental Animals of University of Yamanashi. Transfected melanoma cells (5×10⁶ cells), which were selected by survival after 7 days of blasticidin treatment (Fig. 1B-i, ii), were subcutaneously injected into the dorsal flank skin of mice (Fig. 1B-iii). Formed tumor masses were used for tissue preparation 14 days after the melanoma cell injection (Fig. 1B-iv). IVCT was performed in some mice, as described previously (Ohno et al. 2008). Briefly, all mice were anesthetized by an intraperitoneal administration of pentobarbital sodium (100 mg/kg body weight; Nacalai Tesque, Kyoto, Japan). In three mice, IVCT was performed on exposed subcutaneous tumor masses by directly pouring liquid isopentane-propane (IP) cryogen (-193C), as reported before (Ohno et al. 2008). The frozen tumor tissues were removed with a dental electric drill in liquid nitrogen (-196C) and processed for FS with acetone containing 2% paraformaldehyde. Freeze-substituted tissues were then embedded in paraffin wax for thin sections or OCT compound for cryosections (Fig. 1B-v). Another two mice were transcardially perfused with 2% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) after similar anesthesia, and resected tumor tissues were incubated in the same fixative overnight, dehydrated with graded series of ethanol (Fig. 1B-v), and then embedded in paraffin wax for thin sections.

Hematoxylin-Eosin Staining and Immunostaining

Paraffin sections were serially cut at 4 µm thickness, and mounted on Matsunami Adhesive Slide (MAS)-coated glass slides (Matsunami Glass, Osaka, Japan). They were deparaffinized with xylene and rehydrated in a graded series of ethanol and PBS. Some sections were stained with hematoxylin-eosin (HE). Other deparaffinized sections were incubated with 0.3% hydrogen peroxide in PBS at room temperature for 60 min. They were subsequently incubated with 5% fish gelatin (Sigma-Aldrich) at room temperature for 2 hr and then incubated with a rabbit polyclonal anti-IgM (1/2000) antibody (Bethyl Laboratories, Montgomery, TX) at 4C overnight. Immunocontrol sections were prepared by omitting the primary antibody. The sections were then incubated with a biotinylated goat anti-rabbit IgG (1/500) antibody (Vector Laboratories) at room temperature for 1 hr, and visualized by the avidin-biotin complex-diaminobenzidine (ABC-DAB) method, as described previously (Ohno et al. 2008). Their nuclei were counterstained with 1% methylgreen. Cryosections were cut at 10 or 30 μm thickness for immunofluorescence staining, and were incubated with a goat polyclonal anti-type IV collagen (1/1000) antibody (Southern Biotech, Birmingham, AL) or rabbit polyclonal anti-IgM (1/2000) antibody (Bethyl Laboratories) at 4C overnight, and then counterstained with an AlexaFluor-conjugated donkey anti-goat or anti-rabbit IgG (1/200) antibody (Invitrogen) for 2 hr at room temperature, respectively. The procedures of double-immunostaining for IgM and CAIX or HIF-1α were the same except for the primary antibody: rabbit polyclonal anti-HIF-1α (1/200) antibody (Novus Biologicals, Littleton, CO), rabbit polyclonal anti-CAIX (1/1000) antibody (Novus Biologicals), and goat polyclonal anti-IgM (1/500) antibody (Bethyl Laboratories). The sections were washed in PBS and embedded in Vectashield (Vector Laboratories). Immunostained sections were observed under a light microscope (BX-61; Olympus, Tokyo, Japan) or confocal laser scanning microscope (VT-1000; Olympus). Some cryosections were embedded in Vectashield without immunostaining to check the non-specific fluorescence background.

Image Analyses

All digital images were handled with Fiji (http://fiji.sc/Fiji) and Adobe Photoshop (Adobe Systems, San Jose, CA). In non-anoxic tissues prepared with IVCT, naturally open blood vessels were captured and identified as round/oval profiles that were clearly surrounded by a thin endothelium and also contained groups of flowing erythrocytes with various shapes. Necrotic tissue areas were characterized by the presence of fragmented hematoxylin-, methylgreen-, or DAPI-positive nuclei, and were often accompanied by leaky IgM immunostaining outside blood vessels. Blood vessel diameters were measured along the shorter axis in the transverse sections of oval blood vessel profiles. Distances from blood vessels to necrotic tissue areas were measured on each tissue section as the length from the periphery of the blood vessel to the nearest necrotic region. The thickness of CAIX-negative areas was measured as the longest distance between the edge of each blood vessel and the periphery of CAIX-negative region around the blood vessel. Data for statistical analyses were obtained from individual blood vessels and similar data were pooled from the three tumors. Confocal laser scanning microscopic images for morphometric analyses were obtained with a 100× objective lens (NA=1.4) with 1.5–2× zoom, and the optical resolution was 0.1~0.12 μm/pixel for the x–y axis and 0.5 μm/slice for z-steps. Image stacks were filtered and thresholded, and a 3D Object Counter plugin (Bolte and Cordelieres 2006) was used to measure the volumes of mitoDsRed-positive profiles. Single cells were selected from individual blood vessels that were immunopositive and immunonegative for hypoxia markers, and similar data were pooled from the three tumors. Linear regression and statistical analyses (Mann-Whitney’s U-test and Pearson’s correlation coefficient test) were performed using GraphPad Prism 6 (GraphPad Software, La Jolla, CA).

Results

First, to determine the appropriate fixation conditions for the detection of fluorescence signals in cryofixed specimens, we fixed cultured B16-BL6 cells expressing fluorescent proteins using several fixation methods, including (i) conventional immersion fixation with paraformaldehyde (PFA), (ii) ethanol fixation (EtOH), (iii) FS with acetone only (FS-Ace), (iv) acetone containing 2% paraformaldehyde (FS-PFA) or (v) acetone containing 2% paraformaldehyde and 0.5% glutaraldehyde (FS-GA) (Fig. 1A). Two derivatives of DsRed from Discosoma sp., DsRed2 and mCherry, EGFP from Aequorea victoria and photoswitchable Dendra2 from Dendronephthya sp. were tested. EtOH and FS-Ace retained weak fluorescence signals in cultures expressing mitoDendra2 and mCherry (Fig. 2A-2C; insets, 2D-2F). In contrast, fluorescence signals were better preserved in FS-PFA (Fig. 2G-2I, 2M-2P) than in FS-GA (Fig. 2J-2L), and were similar to those in PFA (Fig. 2A-2C). EGFP and mitoDsRed fluorescence signals were also well maintained with FS-PFA, and mitoDsRed was visualized as granular or tubular mitochondria in the EGFP-immunopositive cytoplasm of B16-BL6 cells (Fig. 2Q-2T). Based on these results, we selected mitoDsRed2 and FS-PFA in subsequent analyses to examine the distributions of fluorescent proteins in tumor tissues in vivo.

Figure 2.

Fluorescence signals of expressed proteins in melanoma cells were well maintained after quick-freezing and freeze-substitution (FS) in vitro. (A-L) Two fluorescence signals of mitochondria-targeted Dendra2 (mitoDen2, green) and mCherry (red), expressed in cultured B16-BL6 cells, were well maintained by immersion fixation with 2% paraformaldehyde (PFA; A-C) or quick-freezing followed by FS with acetone containing 2% paraformaldehyde (FS-PFA; G-I), but were not with quick-freezing followed by FS with acetone containing both paraformaldehyde and glutaraldehyde (FS-GA; J-L). The fluorescence signal was markedly reduced by immersion fixation with ethanol (EtOH; A-C, insets) and quick-freezing followed by FS with only acetone (FS-Ace; D-F). (M-P) Fluorescence signals of mitoDen2 (green) and mCherry (red), expressed in cultured B16-BL6 cells, were well maintained by FS-PFA. Arrowhead indicates the zoomed region of the tissue in the inset. (Q-T) The fluorescence signals of EGFP and mitochondria-targeted DsRed2 (mitoDsRed) were also well maintained by FS-PFA. Tumor cells expressing EGFP with (arrows) or without (arrowheads) mitoDsRed were observed, presumably due to varying degrees of transfection efficiency. Nuclei were stained with DAPI (blue). Bars: 100 μm (A, D, G, J) or 20 μm (M, Q).

We first compared the pure morphology of HE-stained tumor masses subcutaneously formed by injecting mitoDsRed-transfected B16-BL6 cells using different preparation methods, including perfusion fixation followed by conventional dehydration (PF-DH; Fig. 3A, 3B) and IVCT-FS (Fig. 3C, 3D). Conventional PF-DH revealed extensive necrotic tissues around large blood vessels, which were surrounded by viable tumor cells with melanin granules (Fig. 3A, 3B). In the same tissues, a lack of erythrocytes in the blood vessels indicated that the perfusion of fixatives was effective in washing out blood components (Fig. 3B). However, the nuclei and cytoplasm of tumor cells appeared to be shrunken and vacuolated, respectively, which indicated that efficient perfusion for good fixation was hampered by the abnormal architecture of the tumor blood vessels (Fig. 3A, 3B) (Carmeliet and Jain 2002). In contrast, in tissues prepared with IVCT, the appearance of the cytoplasm in viable tumor cells was smooth, and these occasionally contained melanin granules (Fig. 3C, 3D). These tumor cells surrounded open blood vessels with flowing erythrocytes, which were maintained with IVCT as described previously (Ohno et al. 2008). To characterize tumor tissues in more detail, we next immunostained serial paraffin sections of tumor tissues prepared with IVCT for IgM, which is a serum macromolecule commonly localized in functional blood vessels (Ohno et al. 2008). Such immunostaining for IgM was detected within blood vessels with flowing erythrocytes as well as in necrotic tissue areas outside blood vessels, which were identified with HE staining (Fig. 3E, 3F). IgM immunoreactivity in blood vessels and necrotic regions was eliminated in immunocontrol sections (Fig. 3E, 3F). Blood vessels immunopositive for IgM were surrounded by type IV collagen-immunopositive basement membranes, as revealed by double-immunofluorescence staining for type IV collagen and IgM in cryosections (Fig. 4A-4D). Type IV collagen immunoreactivity was also observed in the interstitium, which did not surround IgM-immunopositive blood vessels (Fig. 4A-4D). A correlation was observed between blood vessel diameters and the distances to necrotic tissue areas from the blood vessels in cut tissue sections prepared with IVCT-FS (Fig. 4E, Pearson’s correlation test, r=0.758, p<0.05).

Figure 3.

Tissue morphology and blood vessel morphology in the subcutaneously formed tumor masses of melanoma cells are well maintained by IVCT-FS. (A, B) With conventional perfusion fixation-alcohol dehydration (PF-DH), widely open blood vessels (BV) without erythrocytes and tumor cells with melanin granules (arrowheads in B) are observed between the blood vessels and necrotic regions (asterisks), although cellular nuclei are shrunken and the cytoplasm of tumor cells appeared to be vacuolated. (C, D) With IVCT-FS, open blood vessels with flowing erythrocytes (C, inset, arrow) and surrounding viable tumor cells (D, arrows) with melanin granules (C, inset, arrowheads) are observed adjacent to necrotic regions (D, asterisks). (E, F) Serial paraffin sections of tumors prepared with IVCT-FS show that immunostaining for IgM (F) is detected within blood vessels (BV) and at necrotic areas (asterisks). The same area in the immunocontrol section (F, inset) and its DIC image (E, inset) are also shown. Many tumor cells contain melanin granules (arrowheads). Nuclei are counterstained with methyl green (F). HE: hematoxylin-eosin staining. DIC: differential interference contrast. Boxed areas in A, C, and D are shown in B, D, and C inset, respectively. Bars: 50 μm (A, B, D-F, E inset and F inset), 500 μm (C) or 20 μm (C inset).

Figure 4.

(A-D) Double-immunofluorescence staining for collagen type IV (Col IV, red) and IgM (green) in the cryosections of tumor tissues shows that immunoreactivity for IgM is detected in blood vessels (BV) surrounded by Col IV-immunopositive basement membranes (arrows). The Col IV immunostaining is occasionally observed in the interstitium without the surrounding blood vessels (arrowheads). Nuclei are stained with DAPI (blue). DIC: differential interference contrast. (E) A significantly positive correlation is observed between blood vessel diameter and distance from the blood vessel to the necrotic area. Each dot represents a single blood vessel. *p<0.01 (Pearson’s correlation coefficient test). Bars: 50 µm.

We used two hypoxia markers—carbonic anhydrase IX (CAIX) (Kaluz et al. 2003; Loncaster et al. 2001) and hypoxia-inducible factor-1α (HIF1α) (Semenza 2009)—to determine the hypoxic states of the tumor cells surrounding blood vessels in more detail. HIF1α immunoreactivity was largely observed in tumor cell nuclei (Fig. 5A-5D), whereas CAIX immunostaining appeared to be localized in the plasma membranes of tumor cells (Fig. 5E-5H), which is consistent with the findings of a previous study (Sobhanifar et al. 2005). Areas adjacent to tumor blood vessels were immunonegative for both CAIX and HIF1α, and the highest immunoreactivities of CAIX and HIF1α were observed between CAIX- or HIF1α-immunonegative regions and necrotic regions (Fig. 5A-5H) (Sobhanifar et al. 2005). The distance to CAIX-positive regions from each blood vessel was correlated with the diameter of the blood vessel in these tissue sections (Fig. 5I-5K, Pearson’s correlation test, r=0.698, p<0.05).

Figure 5.

Immunostaining for IgM and hypoxia-inducible factor-1α (HIF1; A-D) or carbonic anhydrase IX (CAIX; E-H) in cryosections prepared with IVCT. The regions surrounding BV are immunonegative for CAIX and HIF1 (A-H, arrowheads), and the immunostaining for CAIX and HIF1 is observed between the non-hypoxic regions and necrotic ones (A-H, arrows). (I-J) Immunostaining for CAIX in a cryosection prepared with IVCT shows thick CAIX-immunonegative regions around a BV (arrowheads) and thinner CAIX-negative regions around a smaller BV (arrows). Necrotic regions contain a few fragmented nuclei (asterisks). (K) Significant correlation is observed between blood vessel diameter and the thickness of the CAIX-immunonegative region around the blood vessel. Each dot represents a single blood vessel. *p<0.01 (Pearson’s correlation coefficient test). Bars: 50 μm (A, E, I).

To investigate whether mitochondrial distributions in the cytoplasm of tumor cells were modulated by the local microenvironment within tumor tissues, some tumor tissue masses formed by the subcutaneous injection of mitoDsRed-expressing tumor cells were carefully examined with IVCT-FS. Although the number of mitoDsRed-positive cells after transfection was limited, target cell selection with blasticidin-resistance rendered by vector transfection increased the ratio of mitoDsRed-positive cells in vitro (Fig. 6A-6H). As a result, an injection of blasticidin-selected cells facilitated the fluorescence detection of mitoDsRed-positive cells in the tumor tissues. Such fluorescence signals of mitoDsRed could also be maintained in cryosections of tumor tissues after FS (Fig. 6E-6H). Although the total sizes of mitoDsRed in tissues were significantly affected by the number of mitoDsRed-positive cells, which varied significantly both in necrotic and non-necrotic regions, the sparse labeling of mitochondria in some tumor cells enabled us to observe mitochondria in individual tumor cells (Fig. 6E-6H). MitoDsRed-positive cells were scattered throughout tumor tissues, which indicated that mitoDsRed expression was maintained in different local microenvironments (Fig. 6G, 6H). When individual tumor cells were observed under confocal laser scanning microscopy, the distributions of mitochondrial networks around each DAPI-positive nucleus could be observed in three dimensions (Fig. 6I-6L).

Figure 6.

The mitochondria-targeted DsRed (Mito) fluorescence signal of melanoma tumor cells was detected in cryosections of tumor tissues prepared by IVCT-FS. (A-D) Although only a fraction of tumor cells expresses Mito fluorescence signals after vector transfection (A, B), all tumor cells expressed these signals following blasticidin drug selection (C, D). (E-H) Mito fluorescence signals were detected in some tumor cells (arrowheads) of tumor tissues (white arrows) with melanin granules (E, black arrows). (I-L) Highly magnified serial images were obtained by confocal laser scanning microscopy (I-K), and three-dimensional reconstruction of the serial optical images of one tumor cell (L) showed interconnected mitochondria around a DAPI-positive (blue) nucleus (arrowheads). Bars: 50 μm (A, C), 20 μm (E) or 5 μm (I, L).

An adequate supply of oxygen was proposed to be limited in tissue areas away from functional blood vessels (Carmeliet and Jain 2002). The distributions of mitoDsRed-positive mitochondria may be modulated in regions distant from IgM-immunopositive blood vessels (Fig. 7A-7D). To investigate whether hypoxia-induced signals could affect mitochondrial volume, which varied significantly, we directly compared the distributions of mitoDsRed-positive mitochondria in cells that did or did not express hypoxia markers, HIF1α (Fig. 7E, 7F) or CAIX (Fig. 7G, 7H.). As described above, the immunoreactivity of HIF1α was primarily found in nuclei (Fig. 7E, 7F), whereas CAIX immunoreactivity appeared to be localized to the plasma membranes of tumor cells and surrounded DsRed-positive mitochondria (Fig. 7G, 7H). We randomly selected HIF1α-negative or -positive cells and CAIX-negative and -positive cells, and measured the total mitochondrial volume of individual tumor cells (Fig. 7I). This analysis revealed that mitochondrial volumes in individual tumor cells significantly decreased in HIF1α-positive or CAIX-positive cells (Fig. 7I). In addition, as shown in Figure 8, fluorescent mitochondria in individual tumor cells were often aligned as cytoplasmic processes extending from cell bodies with DAPI-positive nuclei (Fig. 8A-8E). Double-fluorescence labeling with a component of the basement membranes, type IV collagen (Laurie et al. 1982), revealed that mitochondria were typically localized at the basal side attached to the basement membranes, and that the mitochondria-containing cytoplasmic processes extended along type IV collagen-immunopositive basement membranes (Fig. 8F-8I).

Figure 7.

Mitochondrial volume varied among tumor cells and decreased in tumor cells expressing hypoxia markers. (A-D) Mitochondria-targeted DsRed2-positive cells (Mito, red in C) were scattered throughout tumor tissues and could be observed around blood vessels (arrowheads in C and D) and in necrotic tissue regions (arrows in C and D) immunostained with IgM (green in B and D). *: necrotic regions. DIC: differential interference contrast. (E-H) Single tumor cells expressing mitoDsRed with (F) or without (E) nuclear HIF1α immunostaining and with (H) or without (G) membranous CAIX immunostaining. (I) Scatterplot of total mitochondrial volume in individual tumor cells. The mitochondrial volume of each cell was significantly smaller in tumor cells expressing hypoxia markers than in those not expressing hypoxia markers (**p<0.01, *p<0.05, U-tests). Each dot represents a single cell. n=20 (HIF1α-), 34 (HIF1α+), 16 (CAIX-) or 15 (CAIX+) cells. Lines show the median and interquartile ranges. Nuclei are stained with DAPI (blue). Bars: 20 μm (A) or 10 μm (E-H).

Figure 8.

Some tumor cells extended cytoplasmic processes containing mitochondria along extracellular matrices with immunopositive type IV collagen (Col IV). Serial 1-μm-apart confocal laser scanning microscopic images (A-D) and three-dimensional reconstruction (E) of mitochondria-targeted DsRed2 (Mito, yellow) of a tumor cell located near a blood vessel (BV) showed that relatively small mitochondria were aligned in a cytoplasmic process extending along BV (arrowheads). Asterisks: an endothelial cell. (F-I) Immunostaining for Col IV showed that the mitochondria of many tumor cells localized in local regions attached to Col IV-immunopositive extracellular matrices (arrows), along which Mito-containing processes occasionally extended into tumor tissues (arrowheads). Nuclei were stained with DAPI (blue). Bars: 10 μm (A, E) or 20 μm (F).

Discussion

In the present study, IVCT-FS and subsequent immunohistochemistry were applied to analyses of fluorescence-labeled mitochondria in melanoma cells of living mouse tissues. The fluorescence signals of mitoDsRed2 and other fluorescent proteins were maintained well in specimens prepared with quick-freezing followed by FS. When melanoma cells with fluorescence-labeled mitochondria were subcutaneously injected to form tumor masses and their tissue specimens were prepared with IVCT-FS, the morphology and distributions of mitochondria could be clearly observed in tumor tissues. These mitochondria in individual tumor cells were abundant in local regions that were close to those blood vessels and their volume was significantly decreased by the expression of hypoxia markers. In addition, the distribution of mitochondria was closely associated with the surface of the basement membranes, which is important for tumor cell attachment and invasion, as shown in Figure 8. These findings demonstrate that IVCT in combination with a fluorescence labeling technique is useful for analyses of both the dynamic morphology and subcellular distributions of organelles and proteins in vivo. Furthermore, the volume of mitochondria was dynamically regulated in the hypoxic microenvironment and the distribution of mitochondria was modulated in response to tumor cell interactions with basement membranes.

The constant maintenance of fluorescence probes prepared by immersion fixation and quick-freezing followed by FS depended on features of the chemicals used for the fixation of tissues. As shown in Figure 2, insufficient maintenance of fluorescent probes following ethanol fixation as well as FS with only acetone indicates that such coagulation fixation presumably fails to retain these probes within cells and tissues (Terada et al. 2009). On the other hand, although low concentrations of glutaraldehyde helped to retain small molecules during FS (Terada et al. 2012), diminished fluorescence intensities in response to the addition of glutaraldehyde in the FS solution suggest that chemical glutaraldehyde fixation perturbs the molecular conformation of chromophores to emit fluorescent light (Migneault et al. 2004). Cryofixation followed by FS with an organic solvent containing uranyl acetate was shown to preserve the fluorescent signals of GFP (Nixon et al. 2009). These findings indicate that chemical fixatives of an appropriate strength are critical to maintaining the maximal fluorescence emission in cryofixed tissues. The concentrations of fixatives may need to be optimized for each fluorescent protein in order to achieve the optimal maintenance of fluorescence.

Fluorescent proteins, such as EGFP and mCherry used in the present study, can commonly label specific organelles and subcellular regions by fusing with different targeting protein sequences (Rizzuto et al. 1995). Since fluorescence spectral properties were unchanged after quick-freezing and FS, they were also used in combination with fluorescent dyes such as DAPI. These combined fluorescent probes provide very sensitive pairs for correlative analyses and co-expression experiments with IVCT. Additionally, the genetic arrangement of selection with antibiotic-resistance genes increased the expression levels of fluorescent proteins in the melanoma cells of formed tumor masses and also facilitated the detection of fluorescence-labeled tumor cells. Because of the strong and constant fluorescence signals produced, quick-freezing and IVCT presented in this study will become a useful approach for morphofunctional studies in combination with live-imaging using fluorescent proteins in the cells and tissues of living animals.

In this study, mitochondria in tumor cells were labeled using plasmid-encoding mitoDsRed, and were observed as either small spherical or cylindrical shapes restricted within the cytoplasm of melanoma cells in tumor models in vivo. Previous studies have shown abundant mitochondria with variable and often abnormal morphologies in various types of tumor tissues (Arismendi-Morillo 2009). Metabolic networks including mitochondria appear important even under limited oxygen supply, and mediated by multiple intermediate metabolic pathways that support tumor growth (Schulze and Harris 2012). However, the Warburg effect has been assumed to be involved in the activation of hypoxia-responsive signaling pathways, decreasing mitochondria-dependent energy production along with aberrant mutations in mitochondrial DNA to perturb respiration (Gogvadze et al. 2008; Semenza 2009). In addition, some hypoxic signaling pathways in tumor cells can suppress mitochondrial biogenesis and enhance autophagic degradation of mitochondria to facilitate the mechanism of the metabolic shift from mitochondrial respiration (Sutphin et al. 2007; Zhang et al. 2008). These concepts have been supported by the present result, in which the volumes of mitochondria in melanoma cells were significantly reduced with the expression of hypoxia markers, as shown in Figure 7. HIF-1α is known to be rapidly stabilized or degraded in response to oxygen availability (Jiang et al. 1996), whereas CAIX acts as an indicator of longer hypoxia with slow expression and half-life (Rafajova et al. 2004; Turner et al. 2002). Therefore, these results suggest that mitochondrial biogenesis/degradation may be dynamically regulated by hypoxia to support the metabolism of tumor cells and control their behaviors under different microenvironments. Furthermore, although the results of the present study may be affected by 3-dimentional structures of the tissues, the results of the morphometric analyses performed in the present study suggest that large blood vessels facilitate blood circulation into tumor tissues, and that such widely opened blood vessels may reflect the roles of blood vessel morphology and ischemia/anoxia distribution in the microenvironment of tumor tissues. Further studies using a larger number of samples are needed to reveal specific mechanisms regulating mitochondrial volume in tumor cells with or without hypoxia.

The commonly known modes of migration through extracellular matrices and local invasion across basement membranes are essential for tumor progression and metastasis and are also mediated by interactions between tumor cells and local extracellular matrices (Egeblad et al. 2010; Friedl and Alexander 2011). Mitochondrial localization in such tumor cells is now considered to play important roles in cell-cell interactions (Macaskill et al. 2009; Quintana et al. 2007) and cellular responses to extracellular stimuli in the local microenvironment (Chada and Hollenbeck 2004). Further studies focusing on the mitochondrial behaviors of tumor cells and their processes are necessary to elucidate the morphofunctional roles of mitochondrial localization, especially tumor cell-basement membrane junctions, in tumor cell invasion and metastasis.