Abstract
Adaptive evolution has prompted immune cells to use a wide variety of inhibitory signals, many of which are usurped by tumor cells to evade immune surveillance. Although tumor immunologists often focus on genes and proteins as mediators of immune function, here we highlight two elements from the periodic table—oxygen and potassium—that suppress the immune system in previously unappreciated ways. While both are key to the maintenance of T-cell function and tissue homeostasis, they are exploited by tumors to suppress immuno-surveillance and promote metastatic spread. We discuss the temporal and spatial roles of these elements within the tumor microenvironment and explore possible therapeutic interventions for effective and promising anticancer therapies. Cancer Immunol Res; 5(6); 426–33. ©2017 AACR.
Introduction
Mutations in the genomes of tumor cells generate neoantigens that can be recognized by T cells (1, 2). Despite this immune recognition, growing tumor cells evade immune-mediated destruction to establish primary lesions and to colonize distant and diverse metastatic environments (3–5). Tumors can be targeted with checkpoint modulators or the transfer of T cells against mutated antigens, potentially mediating the complete and durable destruction of tumors (6–8). However, a critical challenge for novel cancer therapies is elucidating mechanisms of immune escape from these therapeutic interventions.
Great strides have been made to uncover the mechanisms of immune suppression that support tumor growth. Such mechanisms include cellular components such as regulatory T cells, immunosuppressive cytokines, intratumoral nutrient availability, and engagement of “checkpoint” molecules such as PD-1 and CTLA-4 (4, 9). Understanding these immune suppression mechanisms has led to the development of new therapeutic agents that show promise in the treatment of multiple cancer histologies. However, it is becoming apparent that both primary and metastatic tumors use multiple resistance mechanisms that vary by tissue type and stage. Also, the spatial distribution of T cells within a tumor influences their function and thus antitumor properties (10). Greater insight into these temporal and spatial regulatory factors may enable the development of new and potent therapeutic strategies for combating tumors.
We and others have undertaken a more fundamental exploration into the elements found in abundance in metazoans. Our research efforts have elucidated how two elements—oxygen and potassium—influence T-cell function, especially within the tumor microenvironment.
Oxygen Is Immunosuppressive for T Cells
Oxygen makes up 65% by mass of the human body. Metastatic tumor cells can colonize healthy tissues that are often very well oxygenated, such as the lungs (5). Sites of metastasis are intimately involved with local vasculature (4). Although this relationship provides a developing metastatic nodule access to nutrient delivery, it also enriches the tumor microenvironment with oxygen. We set out to explore the impact of this abundant element on immunity, hypothesizing that oxygen might inhibit the antitumor T-cell immune response (11).
Like most cells, T cells have an intrinsic capacity to sense oxygen. There are three functionally redundant oxygen sensors in mammals that are members of the prolyl hydroxylase domain containing (PHD) family of proteins. These proteins, which are encoded by homologs of the C. elegans egg-laying-abnormal 9 gene (Egl9), contain iron (Fe), which binds to dioxygen (O2) with great facility (11–13). Cells use these enzymes to signal the presence of O2 by hydroxylating specific proline residues on the hypoxia-inducible factor-α transcription factors (HIFα). The PHD proteins catalyze the addition of one oxygen atom from O2 to proline to create 4-hyroxyproline, whereas the other atom is catalyzed to react with 2-oxoglutarate to form succinate (14).
In the presence of oxygen, hydroxylated proline residues on the α subunits of the HIFs enable them to bind to the von Hippel–Lindau tumor suppressor protein (pVHL), which then directs HIF1α's polyubiquitylation and degradation (15). In the absence of O2 (hypoxia), HIF1α is not degraded and translocates to the nucleus where it forms heterodimers with the constitutively expressed HIF1β. This complex binds to sites on DNA called HIF-responsive elements (HRE), which exist in gene promoters that contain the sequence NCGTG (where N is either A or G). Binding of the HIF1 heterodimer to DNA, complexed with p300, triggers transcription of a variety of genes, many of which control energy metabolism. The functional effects of this oxygen sensing vary by cell type, including among different subsets of T cells (16).
One of the difficulties of studying the PHD oxygen sensors is their redundancy in both mice and humans. In humans, the genes encoding the PHD proteins are on different chromosomes: PHD1 (407 amino acids in length) is encoded by EGLN2, located on chromosome 19q13.2, PHD2 (EGLN1) (426 amino acids) is on chromosome 1q42.1, and PHD3 (EGLN3) (239 amino acids) is on chromosome 14q13.1. This organization is similar in mice, so eliminating PHD protein function in T lymphocytes required a complex breeding scheme. To elucidate the cell-intrinsic role that oxygen sensing plays in influencing T-cell fate and function, we selectively eliminated all three PHD proteins only within the T-cell compartment using the CD4-Cre transgenic mouse model.
By conditionally deleting PHD proteins in the T cells of mice, we found that oxygen sensing by T cells is an important cell-intrinsic mechanism of establishing immunological tolerance in the lungs and other well-oxygenated tissues. Knockout of all three PHD proteins in T lymphocytes (tKO) resulted in spontaneous pulmonary inflammation and made both CD4+ and CD8+ T cells prone to producing the type 1 effector cytokine interferon-γ (IFNγ; Fig. 1; ref. 11). In addition, we observed that the PHD proteins maintain pulmonary-induced T regulatory cells (Treg), which are marked by low levels of the neuropilin-1 (Nrp-1) coreceptor (17–19). Thus, tKO mice have high ratios of cells that produce IFNγ to the Tregs that limit their function (Fig. 1; ref. 11).
Mechanism of oxygen-mediated T-cell immunosuppression and therapeutic interventions to restore the antitumor effector functions of T cells. Under normoxic conditions, PHD proteins function redundantly as oxygen sensors and promote the downregulation of HIF1α levels within T cells. This oxygen-sensing function of the PHD proteins subdues Th1 cell differentiation (T-bet+ cells) and effector activity, concomitantly promoting induced Treg differentiation (Foxp3+ cells). Under hypoxic conditions, T cell effector functions and Th1 differentiation are increased at the expense of decreased iTreg differentiation. The oxygen-mediated immunosuppression of T-cell function under normoxic conditions can be reversed by using small molecule inhibitors like DMOG to block PHD protein activity or through genetic deletion of PHD proteins in T cells using the CRISPR/Cas9 technology.
Just as immune cells in the gut must not overreact to the novel antigenic material to which they are continually exposed (20), there is an evolutionary imperative to be able to breathe in innocuous materials such as smoke, dust, and pollen, which have been present in the environment throughout our evolutionary history. It is also important that immune cells in the lungs and airways are not hyperresponsive to infectious agents such as bacteria and viruses (21). The requirement for moderation of pulmonary immune response derives from the structure of the airways, which tolerate very limited swelling and inflammation before they become obstructed. Bronchiolitis and asthma vividly illustrate the consequences of airway hyper-reactivity (22, 23). Several studies have outlined T-cell–extrinsic mechanisms in which specialized lung cell subsets prime T cells in a manner that limits hyperinflammation, especially in response to innocuous environmental antigens (21, 24, 25). We found that T cells limit their inflammatory response and support the local induction of immunosuppressive Tregs by sensing a highly oxygenated environment through the PHD proteins. The PHD proteins directly connect environmental oxygen with T-cell differentiation and function to support appropriate physiology in well-oxygenated tissues.
The suppression of T-cell effector function by oxygen is evolutionarily adaptive, in that it restrains T-cell inflammation against innocuous environmental antigens in the lung. This same mechanism may also limit beneficial inflammatory responses against colonizing tumor cells, establishing an environment that is favorable for tumor growth and metastasis. As hypothesized, knocking out all three PHD oxygen sensors in T cells significantly reduced pulmonary metastasis. Following intravenous injection of B16 melanoma tumors in tKO mice, the animals developed 3- to 4-fold fewer pulmonary metastases than wild-type mice. Metastatic colonization of the lungs by B16 melanoma tumors incites a local increase in immunosuppressive regulatory T cells that was absent in the tKO mice. In contrast, the growth of subcutaneous tumors was not affected by the loss of PHD sensors (11), suggesting that non-oxygen mechanisms of immunosuppression are dominant in supporting tumor growth in these environments.
Pharmaceutical companies are actively working on inhibitors of PHD proteins to treat anemia. Some of these, including roxadustat, vadadustat, and molidustat, are currently in clinical trials (26, 27). It might be possible to systemically modify or eliminate PHD genes using CRISPR/Cas9 or other emerging technologies, but the risks of eliminating PHD function in all T cells to increase antitumor immune surveillance is likely to be high. Normal mice react to antigens that are innocuous in most people, such as house dust mite (28), by producing a muted immune response that is predominantly characterized by IL4 and IL5 and IL13, a “type-2” or Th2 response. In sharp contrast, tKO mice react to house dust mite antigens with a profound Th1 response characterized by the release of IFNγ in the lungs and produce bloody bronchoalveolar lavage fluid. For tKO mice, challenge with innocuous antigens can be fatal (11).
To summarize, the problem is that T cells with properly functioning PHD proteins permit tumor colonization in the lungs, but the absence of PHD proteins makes T cells hyperresponsive to innocuous antigens. Thus, the global inhibition of PHD proteins using systemic small molecule inhibitors could result in unacceptable immune-mediated toxicities. Additionally, systemically delivered PHD inhibitors could act intrinsically on tumor cells to potentially support their proliferation, as unopposed HIF activity can promote tumor angiogenesis and proliferation (26). One solution to this conundrum is to “drug” PHD proteins only in T cells specific for tumor antigens, while leaving all other T cells intact.
To block PHD activity in tumor-specific T cells, we treated T-cell receptor (TCR) transgenic T cells in vitro with an inhibitory drug called dimethyloxalylglycine (DMOG) that structurally resembles one of the main substrates of the PHD enzymes, 2-oxoglutarate (11). Gene set enrichment analysis (GSEA) revealed that DMOG's effect on gene expression is similar that seen in tKO mice (11), indicating that the immunological effects of DMOG are based on its ability to inhibit the oxygen-sensing PHD proteins. Inhibition of PHD proteins with DMOG prior to adoptive cell transfer–based immunotherapy dramatically changed the functionality of antitumor T cells in a CD4+ model of antitumor immunity, enabling them to produce more IFNγ. DMOG-treated cells also resisted acquiring expression of Foxp3, which encodes a transcription factor that establishes regulatory T-cell ontogeny and coordinates immunoregulatory functional programs (Fig. 1; ref. 29). We have previously shown that regulatory T cells can diminish the functionality of effector T cells in adoptive cell transfer immunotherapy (30, 31).
DMOG-cultured antitumor T cells are significantly better at controlling pulmonary metastases and clearing large, established subcutaneous tumors, prolonging the survival of tumor-bearing mice. Inhibiting the function of the PHD proteins with DMOG also prevents Foxp3 expression in human T cells cultured in standard conditions of 20% oxygen (room air; ref. 11), suggesting that this strategy may improve the effectiveness of adoptive cell transfer for cancer immunotherapy.
Thus, T-cell oxygen-sensing sustains normal pulmonary immune homeostasis in the healthy state, but oxygen sensing also enables tumor metastasis. This is a clear example of tumors “hijacking” normal immunological physiology to support their needs. We predict that this oxygen-driven immunosuppression program supports tumor growth during the earliest stages of tumor development and metastatic establishment, when oxygen is abundantly available in the tumor microenvironment. As tumors grow, however, they become increasingly hypoxic, effectively losing this oxygen-mediated immune evasion. In fact, hypoxia can enhance the proliferation and function of effector memory T cells (32). At this stage, tumors use other mechanisms to suppress T cell–driven antitumor responses. When tumors reach this stage, they experience changes in local concentrations of another element potassium (K+), which has an unexpectedly profound ability to regulate immunity.
Potassium Suppresses T-cell Function in Tumors
Physiologic tissue homeostasis is maintained by a relative equilibrium between cell proliferation and nutrient availability. In contrast, established tumors are characterized by low nutrient availability and rapidly dividing cells, leading to high local levels of cellular apoptosis (“programmed” cell death) and necrosis (lytic cell death; Fig. 2). Though it seems paradoxical, tumor necrosis is an indicator of poor patient survival independent of tumor size and stage (33–35). We sought to explore the mechanisms by which necrosis and the resulting release of necrotic cell contents affect the function of T cells in such microenvironments.
Representative hematoxylin-and-eosin-stained sections of colon cancer and melanoma human tissue sections of patients from the NIH clinical center showing tumor cells, stromal tissue, and necrosis. As seen in these images, the tumor microenvironment can contain highly necrotic regions.
When cells become necrotic, their plasma membranes become increasingly permeable, allowing intracellular contents to egress into the extracellular space. With careful orchestration and at a high energetic expense, cells maintain an electrochemical gradient between the intracellular and extracellular space, via active and passive transport of cations and anions. Mammalian cells are characterized by high concentrations of intracellular potassium (K+), in which the [K+] in the extracellular space ≈ 3–5 mmol/L, but the internal [K+] ≈ 145 mmol/L. The gradient for the sodium ion (Na+) is reversed, with [Na+] being high extracellularly and low within the cell.
We have assessed how the release of cellular contents during necrosis impacts the concentration of ions within the tumor interstitial fluid. Concentrations of the most abundant intracellular ion, K+, are significantly elevated in the tumor interstitial fluid, compared with serum and benign tissue in both mouse and human tumors, with [K+] in the interstitial fluid being ≈ 40 mmol/L, or 5 to 10 times the range of [K+] found in normal serum. This is not the case for other major ions that we measured: Na+, Cl−, Ca2+, and Mg2+. The density of cellular degradation within tumors tissue was directly correlated with the [K+] of the interstitial fluid. Specific experimental induction of either apoptosis or necrosis similarly led to the release of intracellular K+ into the extracellular space (36–38). Similarly, rapid chemotherapy-induced lysis of hematologic malignancies in humans leads to a “tumor lysis syndrome” with associated elevations in serum potassium (39). “Solid tumor lysis syndrome” has also been noted to produce potassium release, with an appreciable elevation of the ion in patient serum (40).
Given that mouse and human tumors contain dense areas of cell necrosis that produce elevations in [K+] within the extracellular space, we investigated how this increase in [K+] in tumor interstitial fluid impacts T-cell function. The additional 40 mmol/L of K+ acutely inhibited TCR-induced production of effector cytokines by T cells. This K+-induced immunosuppression is nonredundant to the coinhibitory signaling of CTLA-4 and PD-1 receptor ligation. A broader characterization of this phenomenon via whole transcriptome RNA-sequencing revealed that elevated extracellular [K+] broadly represses TCR-induced effector programs. Gene-set enrichment analysis (GSEA) showed that genes induced by NF-κB activation or involved in escape from anergy, the adaptive immune response, or cytokine pathways are suppressed by high extracellular K+.
As elevated extracellular [K+] acutely suppresses TCR-driven transcriptional events, we explored whether [K+] could affect TCR-induced signal transduction pathways. TCR ligation rapidly induces signaling cascades that lead to both tyrosine- and serine/threonine phosphorylation, along with influx of divalent ions Ca2+ and Mg2+. Suppression of any of these signal transduction processes can lead to a blunted effector T-cell response. Proximal TCR activation result in a localized cascade of phosphorylation on activating tyrosine residues within a cascade of tyrosine kinases, including Zap70, and leading to activation of the enzyme PLCγ1, producing store-operated Ca2+ efflux (SOCE) as well as activation of the MAPK/Erk pathway. Additionally, TCR ligation, most notably when coordinated with CD28 costimulation, induces phosphoinositol-3-kinase (PI3K) activity with subsequent activation of the serine/threonine kinase Akt and the mammalian target of rapamycin (mTOR), which act to induce protein synthesis, anabolic metabolism, and a program of T-cell effector function via dynamic regulation of the transcription factors HIF1α, c-Myc, Foxo1, and Bach2 among others. We could not detect any changes in TCR-induced Ca2+ flux in the presence of 40 mmol/L isotonic hyperkalemia. Additionally, elevated K+ did not affect TCR-associated tyrosine phosphorylation of Zap70, Erk1/2, or PLCγ-1 (36).
Although TCR-induced Ca2+ flux and tyrosine phosphorylation are normal in the presence of high K+ levels, activity within the Akt–mTOR cascade is profoundly inhibited. Elevated extracellular [K+] suppresses TCR induced Akt phosphorylation, subsequently silencing phosphorylation of the serine/threonine residues targeted by Akt on mTOR and the ribosomal protein S6 (36). The activity of TCR-induced cytokine production and Akt signaling can be restored with pharmacologic or genetic inactivation of protein phosphatase 2A (PP2A), a serine/threonine phosphatase known to dephosphorylate, and thereby inactivate Akt (Fig. 3; ref. 36).
Mechanism of potassium-mediated T-cell immunosuppression and therapeutic interventions to restore the antitumor effector functions of T cells. Tumor cells undergo necrotic cell death and release intracellular K+ ions into the tumor interstitial space; T cells in these environments have more K+ entering the cells. High intracellular potassium within the cell might affect the protein phosphatase PP2A, resulting in the inhibition of Akt and the suppression of T-cell effector functions. Decreased T-cell functions due to raised intracellular [K+] can be reversed through the overexpression of the voltage gated K+ channel Kv1.3 (marked in red).
Mechanistically, we found that K+-induced T-cell suppression is primarily dependent on intracellular [K+]. Pharmacologic interventions that lower the intracellular [K+], augmenting TCR-induced cytokine production, implicate T-cell–intrinsic ion transport as critical to effector functions, antitumor immunity, and immunotherapy. Endogenous control of ion transport in T cells is a complex and relatively understudied field of investigation. Chief among the channels found to have an identified role in T-cell function is the channel Kv1.3 (encoded by the gene Kcna3), a voltage-gated potassium channel linked to T-cell lineage specification and antitumor functions (41).
Under specific conditions, local tumor-induced nutrient restrictions reduce the expression and function of Kv1.3 in locally infiltrating T cells, leading to blocked T-cell activation. The function of Kv1.3 channels in tumor-infiltrating T cells from head and neck tumors is reduced by 70% compared with their function in peripheral blood T cells from the same patients (42–45). Complete loss of Kcna3 can cause CD4+ T cells to take on properties of regulatory cells and reduce production of IFNγ and IL17 cytokines (46). In T cells isolated from head and neck tumors, low Kv1.3 expression is associated with low secretion of the cytotoxic protease granzyme B (47).
When extracellular [K+] increased, we measured an accompanying rise in intracellular [K+] within T cells. To counteract the high [K+] coupled with the downregulation of Kv1.3 function that may occur in large hypoxic tumors, we enforced the expression of Kcna3 in T cells, using a retrovirus. In line with our pharmacologic findings, we found that genetic means to ectopically overexpress Kv1.3 reduced intracellular K+ and augmented effector function. Adoptive transfer of T cells with high enforced expression of Kcna3 into mice with B16 melanoma tumors results in highly phosphorylated Akt and S6 kinases and improved release of IFNγ in the tumor microenvironment. Most importantly, enforced expression of Kcna3 improved T-cell–mediated clearance of large, established tumors (Fig. 3; ref. 36).
PP2A activity suppresses T-cell function in the presence of high [K+], but direct exposure of purified PP2A to high [K+] had no effect on its activity. This may indicate the involvement of other intermediate cell intrinsic molecules. For example, several posttranslational modifications and endogenous small molecules and metabolites, such as ceramides, bimetallic cations, sphingoids, phenols, and polyamines, affect PP2A's localization, targeting, and activity (48, 49).
These findings may also shed light on prior observations that changes in [K+] regulate inflammasome activation in macrophages and can control cellular peptide and phospholipid processing (50, 51). More than a decade ago, elegant experiments revealed that the phagocytic activity of neutrophils is mediated through activation of proteases by intra-organelle potassium flux (52). However, a comprehensive understanding of how immune cells, including T cells, delicately regulate their intracellular potassium concentration to tune signaling cascades is still lacking.
An Elemental View of Tumor-Induced Immunosuppression
Tumor immunologists often draw byzantine diagrams involving multiple cytokines, metabolites, and cell types as being potent suppressors of antitumor immune function. Although much of this tumor immunobiology is undoubtedly true, we have yet to examine the influence on immunity by the elements of the periodic table, which comprise a significant portion of the mammalian tissues. We are often said to be carbon-based organisms, but carbon makes up 18.5% of a person by weight, whereas 65% of any individual is oxygen. Although most elemental oxygen is present in the form of water, fats, carbohydrates, proteins, and nucleic acids, we also breathe in O2, which is essential for cellular respiration. The most abundant cation within cells is potassium, which by weight is twice as plentiful in the body as sodium. Both oxygen and potassium play major roles in maintaining cellular and tissue homeostasis, and our findings reveal that tumors exploit these functions to facilitate their growth and progression as well as promote distant seeding and metastasis.
Although various immune cells participate in tumor immunosurveillance and promote tumor cell arrest, tumor-infiltrating lymphocytes recognizing mutated antigens are likely to be the final common pathway for the immune eradication of tumor cells (53). It is an unfortunate reality for patients that T cells often fail, or fall short, in mediating tumor destruction. This can be due to what has come to be known as adaptive resistance (54), but can also be caused by advantageous mutations in any of a variety of genes involved in antigen processing and presentation (55–59), or by defects in pathways involved in interferon signaling (60) within cancer cells. In addition, localized factors in the tumor microenvironment—such as suppressive immune cells, cytokines, metabolites, nutrients, and even elements—can silence the tumoricidal effector activity of intratumoral T cells, allowing tumors to persist and progress (9, 61–65).
We have highlighted two novel mechanisms of adaptive tissue-specific immune suppression, but multiple mechanisms of immune suppression occur simultaneously and/or progressively within the same tissue, including a myriad of mechanisms not discussed here. Within the K+/O2 immune suppressive axis, oxygen-mediated immune suppression promotes seeding of the disseminated tumor cells to distant well-vascularized tissues like the lungs, but as tumors progress, hypoxic regions emerge as tumor cell growth exceeds its vascular supply. Although hypoxia has been reported to promote tissue necrosis, it has also been shown to promote both acute and long-term inhibition of Kv1.3 in T lymphocytes (Fig. 4). Consequently, this may render intratumoral T cells less fit in accommodating the local increase in potassium concentration that occurs secondary to tumor necrosis (42).
The model illustrates the spatial impact on T cells' antitumor functions by the elements oxygen (O2) and potassium (K+) within the tumor microenvironment. At the time of initial tumor growth and metastases, tumors may experience normoxia. However, as tumor cells grow, vasculature may be inadequately developed, resulting in regions of insufficient oxygen delivery and tissue hypoxia, and cellular necrosis, which results in the release of potassium ions (K+). The inset shows the diffusion gradient of oxygen and potassium within the tumor microenvironment. The tumor contains a limited number of tumor infiltrating lymphocytes, and these T cells are differentially regulated by their immediate microenvironment. Within normoxic regions, T-cell immunosuppression occurs through the work of the PHD oxygen-sensing proteins through the downregulation of HIF1α. T cells immunosuppressed by oxygen are marked blue. Under more hypoxic and hostile necrotic tumor microenvironments, the immunosuppression restricted by the element oxygen is lifted, and in such regions, immunosuppression of T cells occurs through the effect of another element K+, which is a byproduct of tumor cell necrosis. T cells immunosuppressed by potassium are marked green. Whereas the effects of K+ can be counteracted by Kv1.3 expression in T cells, the hypoxic tumor microenvironment downregulates the expression of Kv1.3 in these cells, stranding their effector functions.
This brief review is focused substantially on the impact of oxygen and potassium directly on T cells. However, it is important to note that a considerable body of work indicates that oxygen and potassium can also influence many other immune cells, including tumor-associated macrophages, neutrophils, NK cells, B cells, and many other cell types (51, 52, 66, 67).
Tumor immunotherapy has made great strides, but it is curative in only small fractions of patients who have a limited number of tumor histologies, albeit the list is growing. The immunosuppressive tumor immune compartment challenges basic and translational scientists to devise strategies to enhance anticancer T-cell function, tumor clearance, and subsequent patient survival. We hope and expect that knowledge of “Immunology of the elements” will lead to increases in the efficacy of immunotherapies for patients with cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: D. Gurusamy, D. Clever, R. Eil, N.P. Restifo
Writing, review, and/or revision of the manuscript: D. Gurusamy, D. Clever, R. Eil, N.P. Restifo
Acknowledgments
This work was supported by the Intramural Research Program of the Center for Cancer Research, NCI, NIH. We thank Suman K. Vodnala and Ping-Hsien Lee for helpful discussions and Jennifer Michalowski for excellent editorial input on this article. We are grateful to Richard Lee for histologic images of tumors and Erina He and Alan Hoofring for illustrations.
- Received March 9, 2017.
- Revision received March 30, 2017.
- Accepted April 19, 2017.
- ©2017 American Association for Cancer Research.
References
Volume 5, Issue 6
