Introduction
Localized prostate and breast cancers are highly curable, but once metastasized to remote organs, these cancers are inevitably lethal. Consequently, an important goal of treatment is to identify and exploit specific vulnerabilities of the metastatic cell. Another key aim is to predict which tumors are at high risk of metastasis, allowing potentially toxic therapy to be tailored to those most likely to benefit. These goals have been aided by an improved understanding of cancer cell genetic drift during tumor progression, which allows certain cells to acquire independence from supportive factors in the tissue of origin, to migrate into the vasculature, and to survive and grow at foreign sites such as liver and bone. In addition to genetic drivers of the metastatic phenotype [
1‐
5], epigenetic and protein-level changes have been found to establish tumor cell aggressiveness, including episomal transfer of microRNAs to and from adjacent normal cells, reprogramming of the tumor or stroma by factors released from tumor-infiltrating lymphocytes, and alterations in metabolism brought about by tumor hypoxia [
6‐
8]. Energy production from carbon sources is frequently deranged in cancer, and may be associated with changes in the epigenetic state of the cell that promote cell-autonomous increases in tumor aggression (reviewed in [
8]). A thorough search for metastasis-promoting changes in the cancer cell thus necessarily extends to exploration of protein content, function, and location.
In this study, we used an unsupervised method: differential Cell-SELEX (Systematic Evolution of Ligands by EXponential enrichment), to search for proteins distinguishing metastatic from non-metastatic subclones of a single parental prostate cancer cell line, LNCaP. Owing to their unique, sequence-specific tertiary structure, single-stranded nucleic acids (either DNA or RNA) can bind to individual proteins with high specificity and affinity, comparable to those of antibodies. These oligonucleotides, known as aptamers, can be modified for stability in biological fluids, labeled with fluorescent tags, fused to other molecules, and delivered in vivo without inciting an immune response [
9]. Cell-SELEX uses live cells to select aptamers that recognize cellular proteins in their native and functional state. Differential Cell-SELEX applies the same method to identify aptamers that discriminate between two cell types. The ability to screen large numbers (> 2
40) of sequences increases the likelihood of identifying rare or unique surface marker differences.
Here, we report the identification of a novel RNA aptamer (Apt63) that recognizes a plasma membrane feature that is commonly expressed by multiple aggressive prostate and breast cancer cell lines and tumors, but that exhibits low expression or is absent in non-transformed cells and normal tissues. We demonstrate that the aptamer target is the beta subunit of F
1F
o ATP synthase (ATP5B). This protein is a catalytic component of the final enzyme in cellular ATP production by oxidative phosphorylation, and is located on the inner mitochondrial membrane. ATP5B and other components of the F
1F
o ATP synthase complex have previously been identified on the plasma membrane of certain cell and tumor types, where it is referred to as “ecto-ATP synthase”; several studies have shown that the complex is catalytically active in extracellular ATP production [
10,
11]. Various roles have been established for this activity in a few normal cell types, and particularly in angiogenesis, but its significance and function in cancer remain uncertain. Ecto-ATP synthase acts as a ligand for angiostatin and transduces some of its anti-proliferative and anti-angiogenic effects [
12]. Binding to ecto-ATP synthase by angiostatin, membrane-impermeable small molecules and monoclonal antibodies against the ATP5 beta subunit have been shown to promote cell death in a wide range of susceptible cell types, including HeLa, Leishmania, and plant cells ([
13] and citations therein; [
14‐
16]). Several studies have linked expression of surface ATP synthase to more-aggressive and later-stage cancer [
17,
18], suggesting that the activity of this complex on the cell surface may support the survival of these aggressive cells during the transition to metastasis. In this study, we show that Apt63 distinguishes aggressive breast and prostate cancer cell lines from less-aggressive congenic lines, and from non-transformed cells, both human and murine. In vivo, Apt63 binds selectively to ecto-ATP5B-expressing tumors and not to normal adjacent tissue. Functionally, binding of Apt63 to the plasma membrane exerts selective tumor cell killing by inducing translocation of endonuclease G from mitochondria to nucleus, DNA fragmentation, and apoptosis. We show that that Apt63 plasma membrane binding in clinical tissue biopsies is strongly correlated with advanced tumor stage, and as a corollary, that ATP5B expression in primary tumors is predictive of poor metastasis-free and overall survival. We propose that Apt63 may be useful in early recognition and treatment of a novel subset of highly aggressive primary breast and prostate cancers, defined by surface expression of ATP5B.
Materials and methods
Cell lines and cell culture
Human prostate cancer cell lines used in the Cell-SELEX screen were obtained from Dr. Curtis Pettaway [
19]. Human prostate cancer cells (PC-3, PC3-ML, RWPE-1) were generously provided by Dr. Kerry Burnstein (University of Miami)
, and human breast cancer cell lines (MDA-MB-231, MDA-MB-436, MCF7, MCF10) were obtained from ATCC (Manassas, VA), Murine breast cancer cells lines (4T1, 67NR, E0771, E0771.LMB) were the gift of Dr. Barry Hudson (University of Miami). Dissociated primary tumor lines DT28 and DT22 were the generous gift of Dr. D. El-Ashry (University of Minnesota) [
20]. All cell lines were maintained using the suppliers’ protocols and maintained in 37 °C, 5% CO
2 tissue culture incubators. All cell lines were routinely tested for mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza, Walkersville, MD, USA) and an established PCR protocol [
21].
Differential Cell-SELEX
The pool of RNA aptamers used for Cell-SELEX was obtained from a cDNA library with the general template: TCT CGG ATC CTC AGC GAG TCG TCT G-(N40)-CCG CAT CGT CCT CCC TA (where N40 represents 40 random nucleotides). The cDNA library was amplified by PCR and transcribed in vitro using a Durascribe T7 RNA synthesis kit (Lucigene, USA) with nuclease-stable 2ʹ-F-dCTP and 2ʹ-F-dUTP as previously described [
22]. The aptamer library was purified using an RNeasy kit (Qiagen). Both parental and LNCaP-Pro5 (Pro5) cells were used for negative selection, and LNCaP-LN3 (LN3) for positive selection. Each Cell-SELEX cycle consists two rounds of selection, negative and positive. Parental LNCaP cells were used in the first three selection cycles and Pro5 cells in cycles 4–11. At the beginning of each cycle, 1 µg of the aptamer library was added to 450 µL of PBS containing 0.5 mM MgCl2 and 1 mM CaCl
2 (binding buffer) and RNA was refolded by heating 5 min at 67 °C and cooling at room temperature (RT) for 10 min. All cells used in Cell-SELEX were grown in 75T culture flasks with filtered cups (ThermoFisher Scientific). At 75% confluency, cells were briefly washed with PBS and dissociated from the flask by incubation with 3 ml of Trypsin–EDTA (0.25%) without phenol red (ThermoFisher Scientific) in RT for 3 min. 10 ml of growing media was added to halt Trypsin–EDTA reaction. Detached cells were collected into 15 ml conical tubes (Falcon) and centrifuged for 5 min in a 4 °C tabletop centrifuge (Eppendorf) at 600×
g. Cell pellets were resuspended in 5 ml of binding buffer and counted. 2 × 10
5 Pro5 cells were separated into fresh tubes and centrifuged in tabletop centrifuge for 5 min at RT; the cell pellet was then resuspended with the aptamer library and incubated for 10 min at RT on a circular rotator to continuously agitate the cells. Cells were again centrifuged for 5 min at RT, and the supernatant, containing aptamers not bound to the negative selector, was collected and filtered by passing through 0.2 µm Pall Acrodisc® Sterile Syringe Filters with Supor® Membrane (Pall Laboratory). In the positive selection step, 0.5 × 10
5 LN3 cells were first incubated for 10 min at RT with 0.1 mg/ml of yeast tRNA (Sigma) to reduce non-specific RNA binding, then LN3 cells were washed with binding buffer, centrifuged, and the cell pellet resuspended with the filtered supernatant from the negative selection. LN3 cells were incubated for 10 min on a rotator at RT, followed by isolation of total RNA (including bound aptamers) using Trizol reagent (Invitrogen). RNA aptamers were transcribed from total RNA using an aptamer-specific forward primer and a SuperScript® III Reverse Transcriptase reaction (Invitrogen), and amplified from first strand cDNA by standard PCR (95 °C, 5′, 3 × (94 °C 30″, 52 °C 20″, 72 °C 25″), 15 × (94 °C 30″, 54 °C 20″, 72 °C 25″), 72 °C 5). RNA sequences were transcribed from the resulting cDNA pool using a Durascribe T7 kit as detailed above, and entered into the next Cell-SELEX cycle. RNA aptamer pools were sampled at cycles 1, 4, and 11. After cycle 11, aptamer pools were sequenced, aligned, and analyzed to select candidates for further study as previously described [
22,
23].
Generation of aptamers and scrambled sequences for in vitro and in vivo experiments
Selected aptamers and scrambled oligomers were synthesized by Integrated DNA Technologies (IDT, USA) and used as templates for amplification by PCR (95 °C, 5′, 3 × (94 °C 30″, 52 °C 20″, 72 °C 25″), 15 × (94 °C 30″, 54 °C 20″, 72 °C 25″), 72 °C 5′) with the universal forward primer 5′-GGG GGA ATT CTA ATA CGA CTC ACT ATA GGG AGG ACG ATG CGG-3, and the reverse primer 5′-TCT CGG ATC CTC AGC GAG TCG TC-3′ (oligomer sequences are listed in Online Resource 1). RNA sequences were then transcribed by Durascribe T7 kit (Lucigene, USA) following the manufacturer’s protocol. RNA aptamers and scrambled sequences were labeled for in vitro experiments using Silencer™ siRNA Labeling Kit with Cy™3 dye (Thermo Fisher Scientific) and for in vivo experiments using Ulysis Alexa Fluor™ 647 Nucleic Acid Labeling Kit (Thermo Fisher Scientific) following the manufacturer’s recommended protocols.
Fluorescence microscopy for aptamer imaging
Cells were seeded into 35 mm glass bottom dishes (MatTek Corporation, Ashland, MA) at a density of 0.3 × 106 cells per dish, and allowed to grow for 48 h to 60–75% confluence. Cy3-labeled aptamers were added to culture media at a final concentration of 1 nM and incubated with live cells for 30 min in 37 °C in 5% CO2. Following incubation, cells were washed 3 × for 5 min each with PBS and fixed 10 min with 4% paraformaldehyde at RT. After fixation, cells were washed with PBS and counterstained with DAPI (Sigma), 1 µg final concentration for 5 min. To identify membrane co-localization of Cy3-Apt63 and ATP5B antibody, cells were grown on coverslips in 35 mm tissue culture dishes, and stained sequentially with Cy3-Apt63 and AlexaFluor®647 anti-ATP5B antibodies (ab223436, ABCAM), without a permeabilization step. To co-localize Apt63 and ATP5B antibody within mitochondria, cells were treated with 0.05% Triton X-100 for 5 min, washed three times with PBS, then incubated with both the AlexaFluor®647 anti-ATP5B antibody and Cy3-Apt63. Finally, cells were counterstained with DAPI. For some experiments, live cells were first stained with Cy3-Apt63, followed by fixation, treatment with 0.05% Triton X-100, and DAPI counterstaining as described above. Coverslips were mounted with ProLong®Gold antifade reagent (Life Technologies). Fluorescent images were obtained on a confocal microscope (Leica SP5) using a 20x dry objective (Leica PL APO CS).
Aptamer target purification and identification
Apt63 and scrambled sequence (AptScr) were 3′ end-biotinylated using Pierce™ RNA 3′ End Desthiobiotinylation Kit (ThermoFisher Scientific, USA) following the manufacturer’s protocol. LNCaP-LN3 and LNCaP-Pro5 cells were each seeded into 100 mm Petri dishes for 48 h at a density 0.5 × 106 cells per dish and allowed to grow to 60–75% confluency. On the day of the experiment, cells were incubated at RT for 1 h with the desthiobiotin-RNA-Apt63 or -AptScr complexes, allowing aptamer to bind to target. Following binding, cells were washed 3 × for 5 min each in PBS at RT to remove excess unbound RNA-desthiobiotin complexes, and cross-linked by incubation with 1% paraformaldehyde for 2 min. Next, cells were thoroughly washed 3 × for 5 min each in PBS at RT. To separate membranes from intracellular components, cells were incubated in a mild hypotonic lysis buffer containing 1 M Tris–HCl, 5M NaCl, 50 mM MgCl, 0.1M DTT, and protease inhibitor cocktail for 2 min on ice. Immediately following incubation, cells were gently homogenized in a Dounce homogenizer, ten times on ice, mixed with magnetic beads and left overnight at 4 °C to allow capture of the desthiobiotin-aptamer-target hybrid complexes. On the next day, the beads were thoroughly washed with reagents provided in the kit, and target-aptamer complexes eluted with 30 µL of 8M urea for 10 min at 60 °C. The recovered eluates and total cell lysate were separated on 4–20% gradient SDS–PAGE gels (Bio-Rad, USA). Gels were stained using Pierce™ Silver Stain kits. Protein band distributions were compared between LNCaP-LN3 and LNCaP-Pro5 cell lines, and the most enriched band in the LN3 aptamer-target eluate was cut, sequenced by microcapillary LS/MS/MS, and analyzed by SEQUEST software at Taplin Mass Spectrometry Facility (Harvard Medical School, Boston MA). The predicted protein target was verified by 4–20% gradient SDS-PAGE gel electrophoresis and western blot in whole cell lysates and aptamer eluates using ATP5B antibodies (ab170947, ABCAM) with ATP5B recombinant protein as a positive control (ab92235, ABCAM).
In vitro aptamer binding affinity and cytotoxicity assays
We used two independent methods to evaluate Apt63 cytotoxicity in a series of cell lines in vitro: (1) direct visualization of Apt63 cytotoxicity using the IncuCyte® S3 Live-Cell Analysis System, and (2) SYTOX™ Green uptake. Binding affinity of Apt63 to its membrane target was measured using the CellTiter-Glo® system (Promega). Detailed procedures are described in Online Resource 2.
Mouse xenograft models and aptamer cytotoxicity in vivo
All animal experiments were approved by and performed in accordance with the guidelines of the University of Miami Institutional Animal Care and Use Committee. For live visualization of Apt63 tumor uptake and retention in vivo, a prostate xenograft tumor model was used. NOD.CB17-Prkdcscid/J male mice (10 weeks old,
n = 14, The Jackson Laboratory) were injected orthotopically into the right anterior lobe of the prostate with 2 × 10
6 LN3 and 2 × 10
6 Pro5 cells. For Apt63 cytotoxicity in vivo, we used a previously described breast tumor xenograft mouse model [
24]. NOD.CB17-Prkdcscid/J female mice (6–8 weeks old,
n = 39, The Jackson Laboratory) were injected with 10
6 of MDA-MB-231 cells into the mammary fat pad. In both xenograft models, when tumors were palpable, 1 nmol of Alexa Fluor™ 647-labeled Apt63, AptScr, or unlabeled oligonucleotides suspended in 200 µL of PBS was injected into the tail vein as a single injection. Mice were imaged and euthanized at specified time points. Tumors and selected tissues were dissected and processed for further analysis. Detailed procedures are described in Online Resource 2.
Tumor tissue and FFPE human biopsy arrays fluorescent staining and analysis
Xenograft tumors were removed from euthanized mice and immediately frozen or fixed with 10% buffered formalin (VWR, USA), paraffin embedded, and processed. Prostate and breast core tissue microarrays (TMA) were purchased from US Biomax, Inc (Rockville, MD). The detailed staining protocols are provided in as described in Online Resource 2. Cy3-Apt63-stained human breast biopsy microarrays were imaged by fluorescence microscopy on a Virtual Slide Microscope (VS120) for overview images and on a confocal microscope (Leica SP5) for high-resolution images, and scored for visual presence or absence (greater or less than 10% of cells, respectively) of Apt63 membrane-specific labeling. A list of TMAs with patient information, tumor grade, and stage used in this study and assigned Apt63 score for each biopsy is provided in Online Resource 3. Statistical analysis was performed for the Pearson correlation coefficient of aptamer membrane-specific stain vs. histopathological grades and stages.
ATP5B expression datasets and analysis
ATP5B gene expression was analyzed in prostate and breast cancer samples downloaded from Gene Expression Omnibus and from the Genomic Data Commons Portal. Specifically, RNA-seq data in FPKM (Fragments Per Kilobase Million) and clinical information of the TCGA Prostate Adenocarcinoma dataset (TCGA-PRAD [
25]) were downloaded from the Genomic Data Commons Portal using functions of the
TCGAbiolinks R package and used as is. Expression levels of prostate tumors (
n = 264) and normal prostate tissue samples (
n = 160) from Penney et al. [
26] were downloaded from GEO GSE62872 as Series Matrix File and used as is. Raw data of 545 formalin-fixed paraffin-embedded (FFPE) tissue samples from primary prostate cancer were downloaded from GEO GSE46691 [
27]. Probe-level signals were converted to expression values from CEL files using robust multi-array average procedure RMA [
28] and an Entrez gene-centered custom CDF for Affymetrix Human Exon 1.0 ST Array (
http://brainarray.mbni.med.umich.edu/Brainarray/Database/CustomCDF/CDF_download.asp; version 22). Gene expression profiles of 25 matched normal and tumor breast tissues were downloaded from GEO GSE109169 [
29] as Series Matrix File and used as is. Full expression median-centered data, consisting of 522 primary tumors, 3 metastatic tumors, and 22 tumor-adjacent normal samples, and clinical information of the TCGA Breast Invasive Carcinoma dataset (TCGA-BRCA;[
30]) were downloaded from
https://tcga-data.nci.nih.gov/docs/publications/brca_2012/ and used as is. Finally, we used a breast cancer compendium created from a collection of 4640 samples from 27 major datasets containing microarray data on breast cancer samples annotated with clinical information. The compendium consists of a meta-dataset of gene expression data for 3,661 unique samples from 25 independent cohorts [
31,
32].
All data analyses were performed in R (version 3.5.1) using Bioconductor libraries (BioC 3.7) and R statistical packages. To identify two groups of tumors with either high or low ATP5B expression, we used the classifier described in [
33], based on the standardized expression (score) of a gene or a signature. Tumors were classified as ATP5B ‘Low’ if the ATP5B score was negative and as ATP5B ‘High’ if the ATP5B score was positive. To evaluate the prognostic value of the ATP5B score, we used the Kaplan–Meier method to estimate the probability of metastasis-free survival. To confirm these findings, the Kaplan–Meier curves were compared using the log-rank (Mantel–Cox) test. P-values were calculated according to the standard normal asymptotic distribution, using a cutoff of 0.05 for significance. Survival analysis was performed in GraphPad Prism.
Discussion
Here we show that ectopic plasma membrane ATP5B, a subunit of F
1F
o-ATP synthase, denotes a high metastasis-risk phenotype in breast and prostate cancer, and a vulnerability of cancer cells
in vivo. F
1F
o ATP synthase is a highly conserved enzyme complex residing on the inner mitochondrial membrane, where it conducts the final step in oxidative ATP production. Its 30 protein components are organized into two domains, the F
o proton-translocating domain and the F
1 catalytic domain [
44,
45]. Three pairs of ATP5A and ATP5B subunits form the catalytic core of F
1 in the inner mitochondrial membrane, generating ATP molecules as H + transits the F
o pore. Defects in ATP synthase contribute to diseases including microbial infection, immune deficiency, neuropathies, obesity, diabetes, and cancer [
35,
46,
47].
A plasma membrane-located ATP synthase (ecto-ATP synthase) was initially discovered as a cancer neoantigen more than 20 years ago [
48]. Fully functional ATP synthase complexes have been identified on the plasma membrane of certain normal and many tumor cells, and may either hydrolyze or synthesize ATP [
10,
11]. Ecto-ATP synthase has been proposed to act as a receptor for apo-A1 and thereby to regulate HDL uptake by hepatocytes [
49,
50] and to promote endothelial progenitor cell proliferation and angiogenesis [
51]. Angiostatin has been shown to bind to ecto-ATP synthase and disrupt its ATP synthetic activity, contributing to its anti-angiogenic effects [
10,
52]. However, angiostatin is able to exert these functions through other receptors on the cell surface, including c-met [
53], proteoglycan NG2 [
54], and annexin II [
55]. The importance of these functions of ecto-ATP synthase in normal cells remains to be fully elucidated [
38,
56].
ATP5B emerged in our unbiased screen as a plasma membrane feature that distinguishes the aggressive LNCaP-LN3 cell line from isogenic LNCaP and LNCaP-Pro5 cells, which metastasize infrequently [
19]. Collectively, our data suggest that acquiring this feature may have enabled the metastatic phenotype of the LN3 subclone. Despite substantial effort, no other specific drivers of the aggressive LN3 phenotype have been identified. LN3 cells grow well in the absence of androgen, but do not have androgen receptor amplification [
57]; LN3 also exhibits higher resistance to apoptosis, associated with up-regulation of anti-apoptotic BCL-2 and down-regulation of BAK and BAX [
57]. LN3s express higher levels of macrophage-inhibitory cytokine-1 (MIC1/GDF15) [
58], the chaperone gp96 [
59] and VEGFA [
60], and have greater tumor vascularity [
59,
60] than non-metastatic LNCaP lines. No genetic differences have been shown to explain these properties, although LN3 displays unique deletions in 16q23–qter and 21q of unknown functional significance [
61] and lacks a missense mutation in PlexinB1 found in parental LNCaP cells that appears to be silent [
62]. Previous proteomic analyses found no features distinguishing LN3 from the less-aggressive isogenic lines [
63]. The same study found that endoplasmic reticulum protein ERp5 is overexpressed and displayed on the plasma membrane of both LN3 and Pro5 cells, demonstrating that cycling of intracellular peptides to the plasma membrane is not a rare event during tumorigenesis [
63].
Considerable evidence links ecto-ATP synthase to aggressive cancer cell growth. Plasma membrane-associated ATP5 subunits, including ATP5B, have been correlated with more-aggressive, larger and more advanced tumors, in multiple cancers including breast, lung, and prostate [
17,
42,
64]. In our breast cancer TMA analysis of Apt63 binding, including biopsies representing 416 subjects, surface ATP5B appears to define a unique subset of highly aggressive breast and prostate cancers, present on 45% of DCIS and 55% of invasive ductal carcinomas, and on almost all (91.3%) lymph node metastases. Apt63 staining did not appear to align with tumor size or hormone receptor status, suggesting that ecto-ATP5B denotes an independent tumor phenotype. Surface ATP5B also appears to be important as a tumor-specific survival factor: cancer cells expressing ecto-ATP5B were rapidly killed by Apt63 binding, undergoing nuclear translocation of endonuclease G and DNA fragmentation, while adjacent normal tissues were spared. This selective toxicity could mean that certain breast and prostate tumors are dependent on the presence of functional ecto-ATP synthase, and points to a vulnerability not shared by non-transformed cells.
The biological importance of ecto-ATPase has been explored using a range of physiological and synthetic ligands, including angiostatin, plasminogen, monoclonal antibodies, peptides, and small molecules binding to the F
1 module [
12,
14,
16,
64‐
67]. The effects of these agents are both cell type and ligand-specific, but most reduce extracellular ATP production and cell proliferation, and some initiate programmed death. In HUVECs, which express high levels of ecto-ATP synthase, angiostatin inhibited cell proliferation and ATP production, but was not cytotoxic [
10]; in A549 lung cancer cells, both angiostatin and a polyclonal anti-ATP5B antibody blocked ATP synthesis, induced intracellular acidification, and triggered cell death [
68]. A monoclonal ATP5B antibody (McAb178-5G10) inhibited surface ATP generation and inhibited proliferation of HUVECs and MDA-MB-231 cells, but was not toxic by itself [
69]. The same antibody induced apoptosis in A549 cells, accompanied by falls in extracellular ATP, intracellular pH, and ERK and AKT phosphorylation [
14]. Another monoclonal antibody against ATP5B (mAb6F2C4) inhibited extracellular ATP synthesis, proliferation, anchorage-independent colony formation of the hepatoma cell line SMMC-7721 [
65]; this antibody was also able to reduce hepatoma xenograft growth in vivo. The kringle 1–5 domain of plasminogen, an ecto-ATP synthase ligand, triggered caspase-dependent apoptosis in endothelial cells [
52]. On the other hand, binding of apolipoprotein A1 to ecto-ATP synthase promoted the survival and differentiation of endothelial progenitor cells [
51]. Differences in binding sites, effects on enzyme conformation, and protein interactions of ATP synthase ligands could explain these divergent effects. Additional microenvironmental factors, including acidic extracellular pH, may permit tumor-selective killing [
13,
65]. Further studies will be required to elucidate the specific mechanisms of Apt63-induced programmed cell death in breast and prostate cancer, including effects on extracellular pH, reactive oxygen species, and purinergic nucleotides.
The quantitative relationship between ATP5B gene expression and surface ATP synthase is undetermined and likely complex: ATP synthase subunits are encoded by both nuclear and mitochondrial genomes, and are coordinately regulated through incompletely defined translational and post-translational means [
38,
39,
41,
70‐
72]. Nonetheless, the associations we have identified between ATP5B gene expression and both metastasis-free and overall survival in breast and prostate cancer are remarkable. It is possible that proteomic analysis would demonstrate still stronger links. Comparing the proteomes of MCF-7 breast cancer and a highly invasive subclone, Pan et al. [
42] found that another ATP synthase subunit, ATP5A was overexpressed in the aggressive subclone. ATP5A was identified on the surface of these cells, as well as on MDA-MB-231 and MDA-MB-453 breast cancer cell lines, but not on parental MCF-7 cells, or on non-tumorigenic MCF-10F breast epithelial cells [
42]. In parallel, increased immunoreactive ATP5A was seen in 94% of breast cancers, as well as in 21.2% of normal tissues. This analysis did not discriminate between membrane and cytosolic staining, but the findings are consistent with a relationship between ATP5 protein levels and appearance on the plasma membrane. Other investigators have shown that ATP5B and other subunits of ATP synthase travel on lipid rafts that may shuttle between mitochondrial and plasma membranes [
16,
73]; co-localization with caveolin-1 may be required to maintain a functional surface complex in vascular endothelium [
73]. It will be important to examine the extent to which ATP5B expression correlates with surface ATP5B in future clinical studies of breast cancer prognosis.
A challenge in determining whether and how cancer cells utilize ecto-ATP synthase lies in the essential role of this enzyme in normal cell metabolism, and the unclear pathway by which the complex arrives at the cell surface. Our aptamer represents a new tool that will assist in elucidating these questions. Its rapid and selective cytotoxicity to cells expressing ecto-ATP5B may help to resolve structural and mechanistic questions about the importance of this complex to cancer cell survival and metastasis. Ultimately, the ability of Apt63 ability to target this important but poorly understood tumor antigen in primary breast and prostate tumors may help both to predict and mitigate the risk of future metastasis.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.