III. Cellular ultrastructures in situ as key to understanding tumor energy metabolism: biological significance of the Warburg effect

Animals

Host and graft donor female, athymic, nude mice, 8–9 weeks old, were purchased from Harlan. The donor mouse with GFP-labeled ubiquitin was from The Jackson Laboratory (Stock Number: 004353; Strain Name: C₅₇BL/6-Tg(UBC-GFP)₃₀Scha/J). The mice were housed in the SKCC animal care facility. For surgery, they were anesthetized (7.3 mg ketamine hydrochloride and 2.3 mg xylazine/100 g body weight, inoculated i.p.) and placed on a heating pad. Immediately before tissue harvesting the tumor hosting mice as well as the graft donors were euthanized with pentobarbital overdose (100 mg/kg i.p.).

Cell lines

The parental murine breast cancer cell line, N202.1A²⁵ was stably transfected to express GFP under histone H2B promoter²⁶. The two cell lines, N202.1A parental and N202.1A+H2B-GFP (obtained from Drs. J. Lustgarten and P. Borgstrom) were used to form tumor spheroids by culturing 2×10⁵ cells per well for 2–3 days prior to implantation.

Chambers

A week after establishment of mouse dorsal skin chambers, the tumor spheroids were implanted on a pad of homologous tissue, namely, minced breast fat pad from a lactating mouse (pseudo-orthotopically) or without graft (ectopically) as described earlier²⁷. The tumors were incubated in the chambers for one or three weeks. Their final size was about 1–3 mm in diameter.

Antibodies

The GFP-specific rabbit polyclonal IgG (ab290) was from Abcam; CIB1-specific rabbit polyclonal IgG (11823-1-AP) was from ProteinTech Group, Inc.; CD34-specific rat monoclonal IgG_2a (sc-18917) and non-reactive goat polyclonal IgG (sc-34284) were from Santa Cruz.

Tissue processing

The tumors with some surrounding tissues were dissected out and cut in halves perpendicular to the host skin surface while immersed in cold fixative (4% paraformaldehyde in 0.1 M Na cacodylate pH 7.4). The skin region served later as a reference to distinguish between the edges of the tumor facing the skin and those facing the glass window of the chamber. The halves were then separated and processed independently for TEM and immunocytochemistry.

TEM

The tissues were transferred into a stronger fixative (4% paraformaldehyde/2.0% glutaraldehyde in 0.1 M Na cacodylate pH 7.4) to better preserve the ultrastructures before further cutting. They were cut into 1 mm thick slices in planes perpendicular to the plane of the first cut and to the skin surface, finally, into ~ 1 mm³ blocks, transferred into a fresh portion of the fixative in which they were cut and incubated for 2 hrs at 4°C. The fixed tissue blocks were washed with 0.1 M Na cacodylate – HCl buffer pH 7.4 (3 × 15 min.) and post fixed in 1% OsO₄ in 0.1 M Na cacodylate buffer, pH 7.0 for 60 min. on ice, washed with water and stained with 1% uranyl acetate at RT for one hour. The blocks were embedded in EMbed-12 (EM Sciences, Cat. No 14120). The resin embedded tissues were cut into 60 nm sections, on a Leica Ultracut UCT ultramicrotome and stained with lead citrate²⁸ or viewed without further contrasting.

Immunocytochemistry

During cutting into ~ 1 mm³ blocks as described above, the tissues were kept in the mild fixative to protect antigenic epitopes (4% paraformaldehyde in 0.1 M Na cacodylate pH 7.4). The tissue blocks were vitrified by infiltrating the pieces with 50% PVP (polyvinylpyrrolidone) containing 2.3 M sucrose in 0.1 M Na-cacodylate buffer, pH 7.4, for 2 hrs or overnight, mounted on metal pins and frozen in liquid nitrogen, as described by Tokuyasu²⁹. Frozen tissues were cut into 70 nm sections, on a Leica Ultracut UCT ultramicrotome with the cryo-attachment. The sections were picked from the knife with 2.3 M sucrose and floated on 1% ovalbumin (Sigma, Cat No.A5378) in 0.1 M Na-cacodylate buffer for at least one hour before incubation with specific or non-reactive antibody (50 µg/ml), at RT for one hour. Sections were then rinsed eight times with 0.1% ovalbumin in the same buffer and incubated for one hour with 10 nm Au coupled to protein A (from Dr G. Posthuma; Cell Microscopy Center, University Medical Center Utrecht, The Netherlands). The eight rinsing steps were repeated before fixation of the immune complexes with 1% glutaraldehyde. After rinsing three times with water, the immunostained cryosections were contrasted with a mixture of uranyl acetate and methyl cellulose (25 centipoises, Sigma M-6385) in water, at a final concentration of 1.3% each, for 10 min at RT. Excess liquid was removed and the sections were dried at RT.

Viewing

All sections were viewed and the images captured at 100 kV using a Morgagni 268 electron microscope equipped with a MegaView III digital camera. Images were transmitted from the microscope camera to an iTEM imaging platform from Olympus Soft Imaging Solutions and archived in a designated database. In some cases, the final images were assembled by multiple image alignment (MIA) to increase the surface area without losing the resolution. We used the graphics editing program, Adobe PhotoShop, to add cell type-specific color-coding shown in the twin sets of images included in the Supplement.

Experimental approach

Systems biology research struggles with a paucity of satisfactory methods because local micro-environmental differences, meaningful in vivo, are hard to reproduce in vitro where conditions controlled experimentally affect the analyzed phenomenon indiscriminately. For example, purified DNA molecules can easily be condensed in vitro by dehydration and charge neutralization (adding alcohol and salt, respectively) but the entire length of each molecule is affected simultaneously³³. However in vivo, the process is controlled in a more shrewd way (through protein-DNA interactions), assuring local conformational differences within a single molecule³⁴. Similarly, cells isolated from tissues and cultured in vitro lose their original functional characteristics. To study cellular interactions, preserving tissue structure is necessary. The in situ analysis of ultrathin tissue sections enabled the examination of complex cellular interactions that would have been missed by other methods. It also allowed dialectical interpretation of the observed phenomena that appeared contradictory when studied independently. That approach exposed new relationships between intra- and inter-cellular events and implied logical connections between tissue morphogenesis and metabolic pathways. In vivo, cells with different types of metabolism were located side-by-side and cooperated (Figure 1, Figure 2, Figure S1 and Figure S2). Therefore, in addition to environmental factors (hypoxia and hypoglycemia) that triggered the metabolic switch from oxidative phosphorylation to glycolysis in some cells, tissue-intrinsic regulatory mechanisms must have been involved to control types of metabolism for each cell individually, yet cooperatively for all. The environmental factors triggered the response, but tissue-inherent factors acted as modulators that made the response differential and resulted in changes most beneficial for the tissue rather than individual cells. The in situ analysis demonstrated great heterogeneity of cellular phenotypes within relatively small tissue fragments. Characterizing metabolic pathways of those individual cell types without changing their properties would not have been possible by methods requiring destruction of the tissue fabric. Such relational characterization is necessary to unravel metabolic processes occurring in vivo.

Extramedullar erythrogenesis and tumor cellular heterogeneity

The single piece of information that was most consequential for viewing tumor energy metabolism from a new perspective was the one on the mechanism of erythrogenesis⁵. It showed how the conclusions regarding “erythrocytic enucleation” in other models^35,36 had a long-lasting misleading effect on studies of tumor vasculature morphogenesis. While the presence of non-malignant cells of hematopoietic lineage in tumor samples had been known for years, it was difficult to explain. They were assumed to be “recruited to” or “extravasated at” the tumor site³. However, erythroblasts were not identified in tumors until recently⁵; even though, the morphology of extramedullar erythroblasts (Figure 1, Figure S1, Figure 2, Figure S2, Figure 6[A] and Figure S6[A]) was no different from those in human bone marrow (medulla ossea) seen earlier¹². Previously, the megakaryocyte, erythroblast & EC triplets, as well as fibroblasts, were categorized as stromal cells when seen in tumor nodules. Here, the gradual shrinking and disappearance of mitochondria with a simultaneous proliferation of peroxisomes was correlated with the metabolic switch. That led to a glycolytic pathway in mature erythrosomes through a phase of extensive, peroxisomes-requiring restructuring of the entire cell. That phase lasted long enough to make cells with peroxisomes in the absence of mitochondria commonly encountered. The distinct morphology of the erythroblasts going through the nucleo-cytoplasmic conversion conceptually steered the erythrogenesis process from bone marrow to any location where tissue growth was happening. Practically, it helped in identifying specific locations where vasculature morphogenesis started in vivo (via trans-differentiation of designated cells into HSCs and their interactions with other cells) and in following the process as it unfolded. Clearly, the absence of mitochondria in erythroblasts was a good reason for their respiration to be impaired and replaced by the alternative pathway. That is why using isolated mitochondria to search for causes of the respiratory impairment was not successful, and why metabolic studies on tissue slices were showing both pathways.

Biogenesis of peroxisomes

The nucleo-cytoplasmic conversion that we observed into erythrosomes, being the extreme case of cellular remodeling, augmented the role of peroxisomes in erythrogenesis and helped us to demystify their biogenesis considerably. The process depended on fundamental restructuring of the entire cell in a controlled purposeful fashion. It required reduction of all organelles and membranes to their primary components from which to form erythrosomes - the new organelles with their own specialized membrane. Peroxisomes, known for homing catabolic as well anabolic processes, played a critical role in the restructuring. The results were consistent with the three hypothesized sources of lipids and proteins for the creation of peroxisomal membrane namely mitochondria, ER and cytoplasm²³. However, in cells committed to the erythrogenic autophagy, the nucleus also contributed to the biogenesis of peroxisomes and ultimately erythrosomes (Figure 4[A] and Figure S4[A]).

The mitochondrial reticulum was shown to be present in cultured cells only at certain phase of the cell cycle when the energy output was the greatest, just before synthesis of DNA (G1/S)³¹. It could also be induced by starvation and reactive oxygen species^37,38. Here, we present similar mitochondrial forms in situ, in a tumor cell that was not preparing to replicate because its nucleus was partly degraded and an early EC and potential megakaryocyte accompanied that cell. Together, the three cells represented the typical hematopoietic triplet. The erythrosomal conversion of the nucleated cells into the organelles containing no visible internal ultrastructures (erythrosomes) involved degradation of all organelles originally present in the erythroblasts⁵. The molecular products of such degradation became substrates for the formation of the erythrosomes. The tumor cell in Figure 3 and Figure S3 was in the transition phase and so were its organelles. Mitochondria were converting into peroxisomes, and ultimately the peroxisomes would degrade as well. Hypoxia and starvation could have induced those ultrastructural changes by triggering initiation of erythrosomal authophagy.

The profiles of mitochondria were unlike those fragmented ones in the surrounding tumor cells (eu-chondria). The term meta-chondria describes them better because they were an equivalent of the peroxi-chondria in the secretory cells of glands like the preputial¹⁶, Meibomian³⁹ or those producing gastrointestinal secretions⁴⁰. One can observe that the filamentous mitochondria generating more ATP than any other form³¹ appear when entire cell content requires restructuring, either due to proliferation or conversion of the cell into something different, even if only gradually; for example, the erythrosomes or fat-containing vacuoles. A major similarity between the two cell types harboring such events is that both eventually perish in the process of producing sub-cellular elements: erythrosomes or lipid droplets, respectively. The partly differentiated cells of the mouse preputial gland (type II)¹⁶ were in the process of differentiating into lipid-producing type III and subsequently would degenerate via the lethal cells (type IV)³⁹.

To form peroxisomes, certain elements of mitochondria were utilized, for example those supporting the overlapping functions, including fission of the organelle²², but others had to be synthesized; hence, the involvement of rough ER. The conversion to peroxisomes appeared simultaneous with division of the mitochondrion. The division started by mitochondrion could be completed by peroxisomes. The continuity of partially disassembled mitochondrial membranes with ER (Figure 4[A & B] and Figure S4[A & B]) and localization of histone H2B in the peroxisome (Figure 7[A] and Figure S7[A]) corroborated the close relationship between mitochondria and peroxisomes⁴¹ and provided evidence for direct structural connection in situ. Chromatin was shown here for the first time in peroxisomes, confirming their derivation from mitochondria. A very small amount of mitochondrial DNA found earlier in the preparation of isolated peroxisomes was dismissed as contamination^17,42. Participation of the cytoplasmic components could only be assumed here, except for calmyrites that were usually free in cytoplasm, but in some cells commonly co-localized with peroxisomes and mitochondria (Figure 5, Figure S5, Figure 6 and Figure S6). Substrates catabolized in peroxisomes should be those other than glucose (including amino acids, fatty acids and lactate^43–45), because glycolysis does not require structural isolation from cytoplasm except in some protozoa where specialized peroxisomes are equipped with glycolytic enzymes¹⁴. Therefore, the significance of co-localization of calmyrites with peroxisomes in platelets remains to be clarified.

Mitochondrial conversion into peroxisomes, correlating with hypoxia and cellular differentiation into erythroblasts, reflects concomitant metabolic changes. Erythroblasts and neurons appear to be two extreme examples with regard to their type of metabolism and peroxisomal functions. In the former, peroxisomes are more abundant than mitochondria and eventually replace them, and in the latter, using mainly glucose as a substrate for OXPHOS, peroxisomes are absent¹⁵. In white blood cells, peroxisomes are common and unaccompanied by mitochondria (similar to erythroblasts), indicating that fluctuating oxygen pressure in circulating blood is incompatible with OXPHOS. Some ill-defined granules of white blood cells may represent variants of peroxisomes; a hint for such a possibility was provided by mitochondrial crystalline inter-membrane inclusions in gliomas⁴⁶.

Hemogenic endothelium

According to the concept of “hemogenic endothelium”, blood derives from endothelium³² because the hemogenic embryonal tissue is the tissue from which the aorta develops. The term implies that the endothelium exists before the emergence of blood cells. However, the embryonal “hemogenic endothelium” is not the kind of endothelium present in mature vascular system. As portrayed in gerbil embryo, it morphologically resembled tumor cells rather than the specialized lining of the luminal surface of blood vessels⁴⁷. The essential ultrastructural feature of the embryo shared with the tumor is side-by-side location of cells with differentially modified mitochondria. By contrast, the hallmarks of ECs are a flattened shape, polarized cell membrane (luminal and abluminal), abundant caveolae (mostly on luminal surface), attachment to collagen or basement membrane (on abluminal surface), collagen synthesis and Weibel-Palade bodies. Those bodies have a functional characteristic overlapping with that of alpha granules in platelets, namely involvement in production, storage and secretion of von Willebrand factor which is essential for blood coagulation⁴⁸.

Alternatively, ECs could evolve from hemangioblasts and possibly megakaryocytes and collagen producers (at a higher level of complexity). As with erythrogenesis, differentiation of ECs appears to occur in multiple ways⁵. In our studies, tumor spheroids presented themselves as an analog of undifferentiated embryonic tissue but with a dysfunctional morphogenetic control system, except for vasculature that does make tumors vulnerable to natural selection. Erythrogenesis was a primary event; endothelium evolved next, contrary to the hemogenic endothelium concept, if that concept is taken literally. When comparing the tissues rather than words, one would object not to the “hemogenic” part of the term, but to the word “endothelium”. Yet, under hypoxia, the endothelium could become hemogenic (Figure 8 and Figure S8) just like fibroblasts and tumor cells (Figure 1, Figure S1, Figure 2 and Figure S2). By converting differentiated ECs into erythrosomes, the vessel formation could be disturbed before completion, illustrating how the environment could dominate the intrinsic potential to grow and develop (Figure 8 and Figure S8).

Mini-mitochondria and glycolysis at the mitotic phase of the cell cycle

Glycolysis was detected in mammalian embryos^49,50 when the size of the embryo exceeded the oxygen diffusion range and blood islet appeared (Figure 6(B) in⁵¹), marking the initiation of vasculature development. Activation of glycolysis in non-embryonal proliferating cells has also been known for years, for example in murine splenic lymphocytes in vivo⁵², in cultured rat thymocytes^53–55 or mouse lymphocytes⁵⁶, and in rapidly proliferating prostate cells⁵⁷). The energy required for nuclear division was reported to be wholly or partly derived from the anaerobic metabolism of glucose⁵⁸. The reduced size and the increased electron density of mitochondria at mitosis (mini-chondria) suggests that ATP generation by OXPHOS was down-regulated intermittently, with a frequency determined by the length of the cell cycle (Figure 7[B & C] and Figure S7[B & C]). For that reason, inhibiting glycolysis with the intent to stop vasculature formation could have side effects comparable to chemotherapies targeting cell proliferation.

Biological significance of the Warburg phenomenon

Otto Warburg deserves credit for discovering mammalian glycolysis in 1923 despite it being a minor pathway at tissue level, even if in tissue slices it was enhanced experimentally by cutting off the circulation and providing excess glucose. Warburg was aware of cellular heterogeneity in the tissue slices and thought that solid tumors had mixed metabolic pathways active because of their impurities, i.e. containing non-tumor cells. Concerned with that “contamination”, after 1950 he studied only ascites tumor cells, believing they represented a homogenous population of cells with impaired respiration⁵⁹. Sidney Weinhouse also used tissue slices and ascites tumor cells to study the same metabolic processes but his technique, employing isotopicly labeled metabolites, was much better suited for quantitative analysis and more sensitive in detecting the two pathways, glycolysis and oxidative phosphorylation, in comparison to the Warburg’s manometric method. He debated Warburg’s interpretation repeatedly^1,60–62. The glycolytic process was likely up-regulated in those cells in cyclical, non-synchronous fashion, due to proliferation⁶³ and variable level of hypoxia, depending on the volume of the ascites in vivo, as reflected by morphology of the mitochondria later visualized in ascites tumor cells by TEM^64,65. Here, in situ ultrastructural analysis showed the metabolic differences among neighboring cells, also of the same type. Moreover, it suggested co-existence of the two metabolic pathways even within single cells, based on co-existence of mitochondria and calmyrites either co-localizing with mitochondria or free in the cytoplasm (Figure 5, Figure S5, Figure 6 and Figure S6). The central role of erythrogenesis in vasculature formation deepened our understanding of the relationship between tissue growth and the energy aspect of metabolism by delivering the critical missing element necessary to make the connection between processes occurring at molecular and tissue levels: glycolysis and tissue morphogenesis, including malignancy. Tumors appear to be autistic with regard to interacting with morphogenetic control signals from the system of which they are a part, but capable of basic morphogenetic functions related to vasculature formation; hence their cellular heterogeneity. Otherwise, they would not grow.

The problem with cancer cells is that in many respects they are not different from non-malignant cells. Glycolysis, the less energetically efficient metabolic pathway, appeared critical as a physiological temporary alternative when oxidative phosphorylation diminished due to either growth-related hypoxia or cytokinesis-related slowdown of anabolic processes at the mitotic phase of the cell cycle. It was not tumor specific. Not even yeast could survive for long time on glycolysis alone⁵⁹. The two metabolic pathways were not mutually exclusive. If the biological significance of aerobic glycolysis in vivo was to address the needs of growing tissues⁴, the goal was accomplished indirectly, by inducing vasculature formation and by compensating for cyclically down-regulating mitochondrial activity during mitosis. Regardless of the role that the Warburg effect might play in carcinogenesis⁵⁹, the phenomenon discovered by him appeared to be of profound significance for tissue morphogenesis in general, not only in malignancy. By proposing a fundamental role of metazoan glycolysis in establishing the circulatory system critical for fueling the metabolism of complex organisms, that conclusion indeed brings Warburg’s “anerobic glycolysis” to a renewed prominence, as anticipated by Weinhouse. The essential issues of “regulatory malfunctions which underlie neoplasia” remain unresolved but the level of our thinking about them has changed¹.