Cleveland Clinic Journal of Medicine

Atherosclerosis is recognized as a chronic inflammatory disorder of the vessel wall. Four categories of evidence support the model of atherosclerosis as an inflammatory disease:

• Biomarkers of inflammation are clearly associated with risk and prognosis of atherosclerosis. Three that have been linked conclusively are: C-reactive protein, myeloperoxidase (a marker of leukocyte activation), and antibodies to oxidative modifications of low-density lipoprotein (LDL).
• Tissue studies demonstrate that leukocytes and products of the inflammatory system are prevalent in atherosclerotic plaque.
• Animal models show an absence of atherosclerosis in the absence of monocytes or monocyte recruitment as well as a crucial role for T-cell–derived proinflammatory cytokines.
• It is becoming apparent that patients with chronic systemic inflammatory disorders (eg, systemic lupus erythematosus, Wegener granulomatosis, chronic obesity, and aging) have increased risk of atherosclerosis.

This article examines the mechanisms by which inflammation promotes the development of atherosclerosis and coronary artery disease, with particular attention to the role of CD36, a scavenger receptor for oxidized LDL.

OXIDATION IN PLAQUE FORMATION

Prevailing models that link inflammation to plaque formation suggest that inflammatory stimuli (eg, cigarette smoke, hypertension) provoke changes in the phenotype of the cells of the arterial vessel wall that allow penetration of leukocytes and LDL particles across the endothelial barrier, trapping them in the subendothelial space.^1,2 An inflammatory reaction then occurs in the subendothelial space involving monocytes/macrophages and lymphocytes (especially T cells). Ultimately, through the production and release of oxidizing enzymes such as myeloperoxidase and nitric oxide synthase, the reaction leads to generation of reactive oxygen and nitrogen species. In this setting, LDL particles become modified to a form known as oxidized LDL (oxLDL). OxLDL loses its ability to bind to LDL receptors, which interferes with its normal processing; perhaps more important, oxLDL gains an affinity for a family of proteins called scavenger receptors. Scavenger receptors on macrophages bind and internalize the oxLDL particles, leading to accumulation of cholesterol and other lipids in the cells. Over prolonged periods, increasing quantities of oxLDL become internalized, leading to formation of foam cells (lipid-laden macrophages), the precursor to atherosclerotic plaque. These lipidladen cells are more prone to apoptosis, which further contributes to plaque growth and rupture.

CD36: A CRITICAL SCAVENGER RECEPTOR

One of the most critical scavenger receptors on macrophages is CD36, which is a transmembrane glycoprotein that crosses the membrane twice. CD36 is expressed heavily on monocytes, macrophages, dendritic cells, fat, muscle, capillary endothelial cells, and platelets. It has multiple physiologic functions, including acting as a high-affinity receptor for specific oxidized phospholipids that are found within oxLDL.³ It is also a receptor for phosphatidyl serine (PS) and oxidized PS (oxPS) that is expressed on the surface of apoptotic cells. CD36 is highly conserved in evolution; orthologs are even found in flies, worms, and sponges. Evidence suggests that CD36 and other scavenger receptors probably evolved as part of the innate immune system as recognition molecules for pathogens and pathogen-infected cells.⁴

An interesting aspect of CD36 biology is that its expression on macrophages is increased when the cells are exposed to oxLDL. Among the changes that occur in the lipid components of LDL when it is oxidized is the formation of oxidized fatty acids such as 9- and 13-hydroxy octadecadienoic acid (HODE). These oxidized fatty acids are ligands for the nuclear hormone receptor peroxisome proliferator–activated receptor (PPAR) gamma, a transcription factor that regulates expression of many genes, including CD36. Thus, oxLDL promotes increased expression of CD36 and further cellular uptake of oxLDL. This feed-forward loop presumably accelerates foam cell formation in the arterial neointima. Furthermore, CD36 expression is upregulated at the transcriptional level by inflammatory cytokines such as granulocyte macrophage colony–stimulating factor (GM-CSF), macrophage colony–stimulating factor (M-CSF), and interleukin-4. Hyperglycemia increases CD36 expression through a nontranscriptional mechanism and may contribute to the proatherosclerotic state associated with diabetes.

CD36 mediates atherogenesis

The pathogenic role of oxLDL in atherosclerosis is largely dependent on CD36. Studies using macrophages from genetically altered mice developed in our laboratory that do not express CD36 demonstrated that absence of CD36 expression was associated with a lack of foam cell formation in vitro when cells were exposed to oxLDL. Wild-type mice, in contrast, showed foam cell formation after 12 to 24 hours.⁵

To demonstrate in vivo relevance of these findings, we crossed CD36-null mice with proatherogenic apoE-null mice. When fed an atherogenic Western diet, apoE-null mice develop aortic atherosclerosis within several weeks, in a pattern and histology that closely resembles the human disease. In our experiment, the mice that lacked both CD36 and apoE had a dramatic decrease in the volume of atherosclerosis.⁵ Further studies showed that the proatherogenic role of CD36 was highly likely mediated by the CD36 on macrophages, since transplantation of bone marrow from CD36-null mice into apoE-null mice had the same effect on atherosclerosis as seen in the apoE/CD36-double-null mice.⁶

Scavenger receptor–dependent formation and progression of atherosclerosis is supported by findings of an abundance of oxidized phospholipids that serve as binding partners for CD36 in the plaque region of blood vessels, along with an absence of oxidized phospholipids in the nonplaque region of blood. The enrichment of oxidized phospholipid in the plaque allows CD36 to penetrate the plaque, whereas removal of CD36 drastically decreases the progression of atherosclerosis.

In addition to the formation and progression of plaque, CD36 may be involved in the terminal phases of atherosclerosis (ie, the thrombosis that occurs on a plaque) as a result of abundant CD36 expression on platelets. CD36, in fact, was discovered as a platelet protein and named platelet glycoprotein IV, although for many years the function of CD36 on platelets was not known.

Recent studies by Podrez and colleagues, along with our group, revealed that oxLDL binds to the surface of platelets in a concentration-dependent manner, whereas normal LDL does not. The binding of oxLDL to platelets can be blocked almost completely by inhibiting CD36 with an antibody; binding did not occur with platelets obtained from CD36-deficient mice or people.⁷ Importantly, exposure to oxLDL caused platelets to be activated via a highly specific cell-signaling pathway; low concentrations of oxLDL, such as those found in plasma of individuals with even modest hyperlipidemia, made platelets more sensitive to low doses of “classic” platelet agonists such as collagen and adenosine diphosphate (ADP).⁸ These studies suggest that platelet CD36 could serve as a mechanistic link between inflammation, oxidant stress, and hyperlipidemia to create a prothrombotic state.

It has been known for some time through the work of Eitzman and others that apoE-null mice fed a Western diet are “hypercoagulable”; ie, they have shortened thrombosis times.⁹ This observation led us to investigate the role of CD36 in the hyperlipidemia-associated prothrombotic state. In one experiment, tail-vein bleeding times were measured in apoE-null and apoE/CD36-double-null mice fed a high-fat diet. Whereas the apoE-null animals had markedly shortened bleeding times (~ 2 minutes), the double CD36/apoE-null animals were normal (~ 6–8 minutes).

To examine a model more reflective of pathologic thrombus formation (eg, heart attack, stroke), we induced carotid artery injury in mice by topical application of ferric chloride. This method induces oxidant injury to the endothelium and causes platelet-dependent carotid occlusion. With this model, thrombosis can be monitored in “real time” with a Doppler flow probe and video microscope. As with tail-vein bleeding time, we found that time to carotid occlusion was much shorter in apoE-null mice fed a high-fat diet than in mice fed a normal chow diet or in wild-type mice; further, this prothrombotic state was rescued by genetic ablation of CD36 expression.⁷

Possible role in thrombus formation

More recent experiments from our lab have shown that CD36-null mice fed a normal chow diet have a subtle defect in thrombus formation when arteries or veins are subjected to relatively mild injury.¹⁰ This finding implies a potential role for CD36 in “normal” platelet function and perhaps the existence of an endogenous ligand for CD36 that is unrelated to hyperlipidemia. Since we know that CD36-null mice and CD36-deficient people do not have a bleeding disorder and have normal bleeding times, it is possible that pharmacologic targeting of CD36 may provide a means to inhibit thrombosis without having a major impact on hemostasis.

The possibility that mice or humans can be protected from developing thrombi by blocking CD36 function is supported by initial data obtained from the carotid artery injury model in mice. In the laboratory, an antithrombotic state can be created by blocking the specific CD36-signaling pathway described below. Thrombocytopenic wild-type mice transfused with platelets from wild-type mice exhibit a dramatic increase in thrombosis time when the donor platelets are pretreated with a CD36-signaling inhibitor.⁸ This protective effect vanishes when the same experiment is performed in CD36-null mice.

MICROPARTICLES: MAJOR LIGAND FOR CD36

Based on the experiments described above, we hypothesized the existence of endogenous CD36 ligands involved in thrombosis and proposed that cell-derived microparticles (MPs) were likely candidates. MPs are small (200–1,000 nm) phospholipid vesicles that “bud” off from cells as a result of stimulation or apoptosis. MPs can be derived from endothelial cells, leukocytes, cancer cells, and platelets; they contain selected membrane receptors as well as other proteins inherent to their parental cell (eg, MPs derived from a white cell contain tissue factor that can activate the coagulation cascade). MPs are known to circulate in patients with chronic inflammatory disorders, including acute coronary syndromes, lupus erythematosus, Wegener granulomatosis, and rheumatoid arthritis, and their number probably increases with aging.

Our hypothesis is based on the well-known observation that a major feature of MP generation is a loss of membrane asymmetry; that is, the PS normally expressed on the inner limit of the membrane instead is expressed on the surface. Previous studies from our lab and others had shown that PS and oxPS can be a ligand for CD36.¹¹

To test our hypothesis we developed a rapid flow cytometry assay using an antibody to CD105, an antigen expressed only on endothelial cells, to detect an interaction between endothelial cell–derived MPs and platelets. This interaction is CD36-dependent in that it can be blocked with antibodies to CD36 and does not occur if platelets are taken from mice or humans who are CD36-deficient.¹⁰ MPs behaved like oxLDL in that platelets pretreated with MPs undergo a dramatic augmentation of aggregation in response to low doses of classic agonists. This augmentation of platelet activation does not occur in platelets from CD36-null donors or CD36-null mice.¹⁰

Microparticle accumulation in thrombi

Reprinted from Ghosh A, et al. Platelet CD36 mediates interactions with endothelial cell-derived microparticles and contributes to thrombosis in mice. J Clin Invest 2008; 118:1934–1943. Copyright 2008 by American Society for Clinical Investigation (ASCI).

Specific cytoplasmic signaling cascade

When CD36 binds to its ligands (oxLDL or MPs), it transmits a signal to the cell. In macrophages this signal leads to oxLDL internalization and foam cell formation, while in platelets it contributes to platelet activation and aggregation. In a series of studies from our laboratory and others, it has been shown that these signals are relayed by a series of molecular interactions that involve specific tyrosine kinases from the Src family and serine/threonine kinases from the mitogenactivated protein (MAP) kinase family.¹² The signal to the platelet is mediated by a MAP kinase called c-Jun N-terminal kinase (JNK). Carotid thrombi in wild-type mice stain for the presence of the activated, phosphorylated form of JNK, whereas phospho-JNK expression is decreased by 50% to 60% in carotid thrombi in CD36-null mice,¹² similar to the decrease in MP mass in thrombi from CD36-null mice.

CONCLUSION

These experiments suggest that CD36 has both a proatherogenic and a prothrombotic role in the vascular system. Macrophage CD36 promotes foam cell formation and plaque formation. Platelet CD36 promotes thrombosis by signaling in response to oxLDL and by phospholipids present in cell-derived MP. Therefore, targeting CD36 or CD36-signaling pathways could be a strategy in the treatment of atheroinflammatory disorders and deserves exploration.

Atherosclerosis is recognized as a chronic inflammatory disorder of the vessel wall. Four categories of evidence support the model of atherosclerosis as an inflammatory disease:

• Biomarkers of inflammation are clearly associated with risk and prognosis of atherosclerosis. Three that have been linked conclusively are: C-reactive protein, myeloperoxidase (a marker of leukocyte activation), and antibodies to oxidative modifications of low-density lipoprotein (LDL).
• Tissue studies demonstrate that leukocytes and products of the inflammatory system are prevalent in atherosclerotic plaque.
• Animal models show an absence of atherosclerosis in the absence of monocytes or monocyte recruitment as well as a crucial role for T-cell–derived proinflammatory cytokines.
• It is becoming apparent that patients with chronic systemic inflammatory disorders (eg, systemic lupus erythematosus, Wegener granulomatosis, chronic obesity, and aging) have increased risk of atherosclerosis.

This article examines the mechanisms by which inflammation promotes the development of atherosclerosis and coronary artery disease, with particular attention to the role of CD36, a scavenger receptor for oxidized LDL.

OXIDATION IN PLAQUE FORMATION

Prevailing models that link inflammation to plaque formation suggest that inflammatory stimuli (eg, cigarette smoke, hypertension) provoke changes in the phenotype of the cells of the arterial vessel wall that allow penetration of leukocytes and LDL particles across the endothelial barrier, trapping them in the subendothelial space.^1,2 An inflammatory reaction then occurs in the subendothelial space involving monocytes/macrophages and lymphocytes (especially T cells). Ultimately, through the production and release of oxidizing enzymes such as myeloperoxidase and nitric oxide synthase, the reaction leads to generation of reactive oxygen and nitrogen species. In this setting, LDL particles become modified to a form known as oxidized LDL (oxLDL). OxLDL loses its ability to bind to LDL receptors, which interferes with its normal processing; perhaps more important, oxLDL gains an affinity for a family of proteins called scavenger receptors. Scavenger receptors on macrophages bind and internalize the oxLDL particles, leading to accumulation of cholesterol and other lipids in the cells. Over prolonged periods, increasing quantities of oxLDL become internalized, leading to formation of foam cells (lipid-laden macrophages), the precursor to atherosclerotic plaque. These lipidladen cells are more prone to apoptosis, which further contributes to plaque growth and rupture.

CD36: A CRITICAL SCAVENGER RECEPTOR

One of the most critical scavenger receptors on macrophages is CD36, which is a transmembrane glycoprotein that crosses the membrane twice. CD36 is expressed heavily on monocytes, macrophages, dendritic cells, fat, muscle, capillary endothelial cells, and platelets. It has multiple physiologic functions, including acting as a high-affinity receptor for specific oxidized phospholipids that are found within oxLDL.³ It is also a receptor for phosphatidyl serine (PS) and oxidized PS (oxPS) that is expressed on the surface of apoptotic cells. CD36 is highly conserved in evolution; orthologs are even found in flies, worms, and sponges. Evidence suggests that CD36 and other scavenger receptors probably evolved as part of the innate immune system as recognition molecules for pathogens and pathogen-infected cells.⁴

An interesting aspect of CD36 biology is that its expression on macrophages is increased when the cells are exposed to oxLDL. Among the changes that occur in the lipid components of LDL when it is oxidized is the formation of oxidized fatty acids such as 9- and 13-hydroxy octadecadienoic acid (HODE). These oxidized fatty acids are ligands for the nuclear hormone receptor peroxisome proliferator–activated receptor (PPAR) gamma, a transcription factor that regulates expression of many genes, including CD36. Thus, oxLDL promotes increased expression of CD36 and further cellular uptake of oxLDL. This feed-forward loop presumably accelerates foam cell formation in the arterial neointima. Furthermore, CD36 expression is upregulated at the transcriptional level by inflammatory cytokines such as granulocyte macrophage colony–stimulating factor (GM-CSF), macrophage colony–stimulating factor (M-CSF), and interleukin-4. Hyperglycemia increases CD36 expression through a nontranscriptional mechanism and may contribute to the proatherosclerotic state associated with diabetes.

CD36 mediates atherogenesis

The pathogenic role of oxLDL in atherosclerosis is largely dependent on CD36. Studies using macrophages from genetically altered mice developed in our laboratory that do not express CD36 demonstrated that absence of CD36 expression was associated with a lack of foam cell formation in vitro when cells were exposed to oxLDL. Wild-type mice, in contrast, showed foam cell formation after 12 to 24 hours.⁵

To demonstrate in vivo relevance of these findings, we crossed CD36-null mice with proatherogenic apoE-null mice. When fed an atherogenic Western diet, apoE-null mice develop aortic atherosclerosis within several weeks, in a pattern and histology that closely resembles the human disease. In our experiment, the mice that lacked both CD36 and apoE had a dramatic decrease in the volume of atherosclerosis.⁵ Further studies showed that the proatherogenic role of CD36 was highly likely mediated by the CD36 on macrophages, since transplantation of bone marrow from CD36-null mice into apoE-null mice had the same effect on atherosclerosis as seen in the apoE/CD36-double-null mice.⁶

Scavenger receptor–dependent formation and progression of atherosclerosis is supported by findings of an abundance of oxidized phospholipids that serve as binding partners for CD36 in the plaque region of blood vessels, along with an absence of oxidized phospholipids in the nonplaque region of blood. The enrichment of oxidized phospholipid in the plaque allows CD36 to penetrate the plaque, whereas removal of CD36 drastically decreases the progression of atherosclerosis.

In addition to the formation and progression of plaque, CD36 may be involved in the terminal phases of atherosclerosis (ie, the thrombosis that occurs on a plaque) as a result of abundant CD36 expression on platelets. CD36, in fact, was discovered as a platelet protein and named platelet glycoprotein IV, although for many years the function of CD36 on platelets was not known.

Recent studies by Podrez and colleagues, along with our group, revealed that oxLDL binds to the surface of platelets in a concentration-dependent manner, whereas normal LDL does not. The binding of oxLDL to platelets can be blocked almost completely by inhibiting CD36 with an antibody; binding did not occur with platelets obtained from CD36-deficient mice or people.⁷ Importantly, exposure to oxLDL caused platelets to be activated via a highly specific cell-signaling pathway; low concentrations of oxLDL, such as those found in plasma of individuals with even modest hyperlipidemia, made platelets more sensitive to low doses of “classic” platelet agonists such as collagen and adenosine diphosphate (ADP).⁸ These studies suggest that platelet CD36 could serve as a mechanistic link between inflammation, oxidant stress, and hyperlipidemia to create a prothrombotic state.

It has been known for some time through the work of Eitzman and others that apoE-null mice fed a Western diet are “hypercoagulable”; ie, they have shortened thrombosis times.⁹ This observation led us to investigate the role of CD36 in the hyperlipidemia-associated prothrombotic state. In one experiment, tail-vein bleeding times were measured in apoE-null and apoE/CD36-double-null mice fed a high-fat diet. Whereas the apoE-null animals had markedly shortened bleeding times (~ 2 minutes), the double CD36/apoE-null animals were normal (~ 6–8 minutes).

To examine a model more reflective of pathologic thrombus formation (eg, heart attack, stroke), we induced carotid artery injury in mice by topical application of ferric chloride. This method induces oxidant injury to the endothelium and causes platelet-dependent carotid occlusion. With this model, thrombosis can be monitored in “real time” with a Doppler flow probe and video microscope. As with tail-vein bleeding time, we found that time to carotid occlusion was much shorter in apoE-null mice fed a high-fat diet than in mice fed a normal chow diet or in wild-type mice; further, this prothrombotic state was rescued by genetic ablation of CD36 expression.⁷

Possible role in thrombus formation

More recent experiments from our lab have shown that CD36-null mice fed a normal chow diet have a subtle defect in thrombus formation when arteries or veins are subjected to relatively mild injury.¹⁰ This finding implies a potential role for CD36 in “normal” platelet function and perhaps the existence of an endogenous ligand for CD36 that is unrelated to hyperlipidemia. Since we know that CD36-null mice and CD36-deficient people do not have a bleeding disorder and have normal bleeding times, it is possible that pharmacologic targeting of CD36 may provide a means to inhibit thrombosis without having a major impact on hemostasis.

The possibility that mice or humans can be protected from developing thrombi by blocking CD36 function is supported by initial data obtained from the carotid artery injury model in mice. In the laboratory, an antithrombotic state can be created by blocking the specific CD36-signaling pathway described below. Thrombocytopenic wild-type mice transfused with platelets from wild-type mice exhibit a dramatic increase in thrombosis time when the donor platelets are pretreated with a CD36-signaling inhibitor.⁸ This protective effect vanishes when the same experiment is performed in CD36-null mice.

MICROPARTICLES: MAJOR LIGAND FOR CD36

Based on the experiments described above, we hypothesized the existence of endogenous CD36 ligands involved in thrombosis and proposed that cell-derived microparticles (MPs) were likely candidates. MPs are small (200–1,000 nm) phospholipid vesicles that “bud” off from cells as a result of stimulation or apoptosis. MPs can be derived from endothelial cells, leukocytes, cancer cells, and platelets; they contain selected membrane receptors as well as other proteins inherent to their parental cell (eg, MPs derived from a white cell contain tissue factor that can activate the coagulation cascade). MPs are known to circulate in patients with chronic inflammatory disorders, including acute coronary syndromes, lupus erythematosus, Wegener granulomatosis, and rheumatoid arthritis, and their number probably increases with aging.

Our hypothesis is based on the well-known observation that a major feature of MP generation is a loss of membrane asymmetry; that is, the PS normally expressed on the inner limit of the membrane instead is expressed on the surface. Previous studies from our lab and others had shown that PS and oxPS can be a ligand for CD36.¹¹

To test our hypothesis we developed a rapid flow cytometry assay using an antibody to CD105, an antigen expressed only on endothelial cells, to detect an interaction between endothelial cell–derived MPs and platelets. This interaction is CD36-dependent in that it can be blocked with antibodies to CD36 and does not occur if platelets are taken from mice or humans who are CD36-deficient.¹⁰ MPs behaved like oxLDL in that platelets pretreated with MPs undergo a dramatic augmentation of aggregation in response to low doses of classic agonists. This augmentation of platelet activation does not occur in platelets from CD36-null donors or CD36-null mice.¹⁰

Microparticle accumulation in thrombi

Reprinted from Ghosh A, et al. Platelet CD36 mediates interactions with endothelial cell-derived microparticles and contributes to thrombosis in mice. J Clin Invest 2008; 118:1934–1943. Copyright 2008 by American Society for Clinical Investigation (ASCI).

Specific cytoplasmic signaling cascade

When CD36 binds to its ligands (oxLDL or MPs), it transmits a signal to the cell. In macrophages this signal leads to oxLDL internalization and foam cell formation, while in platelets it contributes to platelet activation and aggregation. In a series of studies from our laboratory and others, it has been shown that these signals are relayed by a series of molecular interactions that involve specific tyrosine kinases from the Src family and serine/threonine kinases from the mitogenactivated protein (MAP) kinase family.¹² The signal to the platelet is mediated by a MAP kinase called c-Jun N-terminal kinase (JNK). Carotid thrombi in wild-type mice stain for the presence of the activated, phosphorylated form of JNK, whereas phospho-JNK expression is decreased by 50% to 60% in carotid thrombi in CD36-null mice,¹² similar to the decrease in MP mass in thrombi from CD36-null mice.

CONCLUSION

These experiments suggest that CD36 has both a proatherogenic and a prothrombotic role in the vascular system. Macrophage CD36 promotes foam cell formation and plaque formation. Platelet CD36 promotes thrombosis by signaling in response to oxLDL and by phospholipids present in cell-derived MP. Therefore, targeting CD36 or CD36-signaling pathways could be a strategy in the treatment of atheroinflammatory disorders and deserves exploration.

Atherosclerosis is recognized as a chronic inflammatory disorder of the vessel wall. Four categories of evidence support the model of atherosclerosis as an inflammatory disease:

• Biomarkers of inflammation are clearly associated with risk and prognosis of atherosclerosis. Three that have been linked conclusively are: C-reactive protein, myeloperoxidase (a marker of leukocyte activation), and antibodies to oxidative modifications of low-density lipoprotein (LDL).
• Tissue studies demonstrate that leukocytes and products of the inflammatory system are prevalent in atherosclerotic plaque.
• Animal models show an absence of atherosclerosis in the absence of monocytes or monocyte recruitment as well as a crucial role for T-cell–derived proinflammatory cytokines.
• It is becoming apparent that patients with chronic systemic inflammatory disorders (eg, systemic lupus erythematosus, Wegener granulomatosis, chronic obesity, and aging) have increased risk of atherosclerosis.

This article examines the mechanisms by which inflammation promotes the development of atherosclerosis and coronary artery disease, with particular attention to the role of CD36, a scavenger receptor for oxidized LDL.

OXIDATION IN PLAQUE FORMATION

Prevailing models that link inflammation to plaque formation suggest that inflammatory stimuli (eg, cigarette smoke, hypertension) provoke changes in the phenotype of the cells of the arterial vessel wall that allow penetration of leukocytes and LDL particles across the endothelial barrier, trapping them in the subendothelial space.^1,2 An inflammatory reaction then occurs in the subendothelial space involving monocytes/macrophages and lymphocytes (especially T cells). Ultimately, through the production and release of oxidizing enzymes such as myeloperoxidase and nitric oxide synthase, the reaction leads to generation of reactive oxygen and nitrogen species. In this setting, LDL particles become modified to a form known as oxidized LDL (oxLDL). OxLDL loses its ability to bind to LDL receptors, which interferes with its normal processing; perhaps more important, oxLDL gains an affinity for a family of proteins called scavenger receptors. Scavenger receptors on macrophages bind and internalize the oxLDL particles, leading to accumulation of cholesterol and other lipids in the cells. Over prolonged periods, increasing quantities of oxLDL become internalized, leading to formation of foam cells (lipid-laden macrophages), the precursor to atherosclerotic plaque. These lipidladen cells are more prone to apoptosis, which further contributes to plaque growth and rupture.

CD36: A CRITICAL SCAVENGER RECEPTOR

One of the most critical scavenger receptors on macrophages is CD36, which is a transmembrane glycoprotein that crosses the membrane twice. CD36 is expressed heavily on monocytes, macrophages, dendritic cells, fat, muscle, capillary endothelial cells, and platelets. It has multiple physiologic functions, including acting as a high-affinity receptor for specific oxidized phospholipids that are found within oxLDL.³ It is also a receptor for phosphatidyl serine (PS) and oxidized PS (oxPS) that is expressed on the surface of apoptotic cells. CD36 is highly conserved in evolution; orthologs are even found in flies, worms, and sponges. Evidence suggests that CD36 and other scavenger receptors probably evolved as part of the innate immune system as recognition molecules for pathogens and pathogen-infected cells.⁴

An interesting aspect of CD36 biology is that its expression on macrophages is increased when the cells are exposed to oxLDL. Among the changes that occur in the lipid components of LDL when it is oxidized is the formation of oxidized fatty acids such as 9- and 13-hydroxy octadecadienoic acid (HODE). These oxidized fatty acids are ligands for the nuclear hormone receptor peroxisome proliferator–activated receptor (PPAR) gamma, a transcription factor that regulates expression of many genes, including CD36. Thus, oxLDL promotes increased expression of CD36 and further cellular uptake of oxLDL. This feed-forward loop presumably accelerates foam cell formation in the arterial neointima. Furthermore, CD36 expression is upregulated at the transcriptional level by inflammatory cytokines such as granulocyte macrophage colony–stimulating factor (GM-CSF), macrophage colony–stimulating factor (M-CSF), and interleukin-4. Hyperglycemia increases CD36 expression through a nontranscriptional mechanism and may contribute to the proatherosclerotic state associated with diabetes.

CD36 mediates atherogenesis

The pathogenic role of oxLDL in atherosclerosis is largely dependent on CD36. Studies using macrophages from genetically altered mice developed in our laboratory that do not express CD36 demonstrated that absence of CD36 expression was associated with a lack of foam cell formation in vitro when cells were exposed to oxLDL. Wild-type mice, in contrast, showed foam cell formation after 12 to 24 hours.⁵

To demonstrate in vivo relevance of these findings, we crossed CD36-null mice with proatherogenic apoE-null mice. When fed an atherogenic Western diet, apoE-null mice develop aortic atherosclerosis within several weeks, in a pattern and histology that closely resembles the human disease. In our experiment, the mice that lacked both CD36 and apoE had a dramatic decrease in the volume of atherosclerosis.⁵ Further studies showed that the proatherogenic role of CD36 was highly likely mediated by the CD36 on macrophages, since transplantation of bone marrow from CD36-null mice into apoE-null mice had the same effect on atherosclerosis as seen in the apoE/CD36-double-null mice.⁶

Scavenger receptor–dependent formation and progression of atherosclerosis is supported by findings of an abundance of oxidized phospholipids that serve as binding partners for CD36 in the plaque region of blood vessels, along with an absence of oxidized phospholipids in the nonplaque region of blood. The enrichment of oxidized phospholipid in the plaque allows CD36 to penetrate the plaque, whereas removal of CD36 drastically decreases the progression of atherosclerosis.

In addition to the formation and progression of plaque, CD36 may be involved in the terminal phases of atherosclerosis (ie, the thrombosis that occurs on a plaque) as a result of abundant CD36 expression on platelets. CD36, in fact, was discovered as a platelet protein and named platelet glycoprotein IV, although for many years the function of CD36 on platelets was not known.

Recent studies by Podrez and colleagues, along with our group, revealed that oxLDL binds to the surface of platelets in a concentration-dependent manner, whereas normal LDL does not. The binding of oxLDL to platelets can be blocked almost completely by inhibiting CD36 with an antibody; binding did not occur with platelets obtained from CD36-deficient mice or people.⁷ Importantly, exposure to oxLDL caused platelets to be activated via a highly specific cell-signaling pathway; low concentrations of oxLDL, such as those found in plasma of individuals with even modest hyperlipidemia, made platelets more sensitive to low doses of “classic” platelet agonists such as collagen and adenosine diphosphate (ADP).⁸ These studies suggest that platelet CD36 could serve as a mechanistic link between inflammation, oxidant stress, and hyperlipidemia to create a prothrombotic state.

It has been known for some time through the work of Eitzman and others that apoE-null mice fed a Western diet are “hypercoagulable”; ie, they have shortened thrombosis times.⁹ This observation led us to investigate the role of CD36 in the hyperlipidemia-associated prothrombotic state. In one experiment, tail-vein bleeding times were measured in apoE-null and apoE/CD36-double-null mice fed a high-fat diet. Whereas the apoE-null animals had markedly shortened bleeding times (~ 2 minutes), the double CD36/apoE-null animals were normal (~ 6–8 minutes).

To examine a model more reflective of pathologic thrombus formation (eg, heart attack, stroke), we induced carotid artery injury in mice by topical application of ferric chloride. This method induces oxidant injury to the endothelium and causes platelet-dependent carotid occlusion. With this model, thrombosis can be monitored in “real time” with a Doppler flow probe and video microscope. As with tail-vein bleeding time, we found that time to carotid occlusion was much shorter in apoE-null mice fed a high-fat diet than in mice fed a normal chow diet or in wild-type mice; further, this prothrombotic state was rescued by genetic ablation of CD36 expression.⁷

Possible role in thrombus formation

More recent experiments from our lab have shown that CD36-null mice fed a normal chow diet have a subtle defect in thrombus formation when arteries or veins are subjected to relatively mild injury.¹⁰ This finding implies a potential role for CD36 in “normal” platelet function and perhaps the existence of an endogenous ligand for CD36 that is unrelated to hyperlipidemia. Since we know that CD36-null mice and CD36-deficient people do not have a bleeding disorder and have normal bleeding times, it is possible that pharmacologic targeting of CD36 may provide a means to inhibit thrombosis without having a major impact on hemostasis.

The possibility that mice or humans can be protected from developing thrombi by blocking CD36 function is supported by initial data obtained from the carotid artery injury model in mice. In the laboratory, an antithrombotic state can be created by blocking the specific CD36-signaling pathway described below. Thrombocytopenic wild-type mice transfused with platelets from wild-type mice exhibit a dramatic increase in thrombosis time when the donor platelets are pretreated with a CD36-signaling inhibitor.⁸ This protective effect vanishes when the same experiment is performed in CD36-null mice.

MICROPARTICLES: MAJOR LIGAND FOR CD36

Based on the experiments described above, we hypothesized the existence of endogenous CD36 ligands involved in thrombosis and proposed that cell-derived microparticles (MPs) were likely candidates. MPs are small (200–1,000 nm) phospholipid vesicles that “bud” off from cells as a result of stimulation or apoptosis. MPs can be derived from endothelial cells, leukocytes, cancer cells, and platelets; they contain selected membrane receptors as well as other proteins inherent to their parental cell (eg, MPs derived from a white cell contain tissue factor that can activate the coagulation cascade). MPs are known to circulate in patients with chronic inflammatory disorders, including acute coronary syndromes, lupus erythematosus, Wegener granulomatosis, and rheumatoid arthritis, and their number probably increases with aging.

Our hypothesis is based on the well-known observation that a major feature of MP generation is a loss of membrane asymmetry; that is, the PS normally expressed on the inner limit of the membrane instead is expressed on the surface. Previous studies from our lab and others had shown that PS and oxPS can be a ligand for CD36.¹¹

To test our hypothesis we developed a rapid flow cytometry assay using an antibody to CD105, an antigen expressed only on endothelial cells, to detect an interaction between endothelial cell–derived MPs and platelets. This interaction is CD36-dependent in that it can be blocked with antibodies to CD36 and does not occur if platelets are taken from mice or humans who are CD36-deficient.¹⁰ MPs behaved like oxLDL in that platelets pretreated with MPs undergo a dramatic augmentation of aggregation in response to low doses of classic agonists. This augmentation of platelet activation does not occur in platelets from CD36-null donors or CD36-null mice.¹⁰

Microparticle accumulation in thrombi

Reprinted from Ghosh A, et al. Platelet CD36 mediates interactions with endothelial cell-derived microparticles and contributes to thrombosis in mice. J Clin Invest 2008; 118:1934–1943. Copyright 2008 by American Society for Clinical Investigation (ASCI).

Specific cytoplasmic signaling cascade

When CD36 binds to its ligands (oxLDL or MPs), it transmits a signal to the cell. In macrophages this signal leads to oxLDL internalization and foam cell formation, while in platelets it contributes to platelet activation and aggregation. In a series of studies from our laboratory and others, it has been shown that these signals are relayed by a series of molecular interactions that involve specific tyrosine kinases from the Src family and serine/threonine kinases from the mitogenactivated protein (MAP) kinase family.¹² The signal to the platelet is mediated by a MAP kinase called c-Jun N-terminal kinase (JNK). Carotid thrombi in wild-type mice stain for the presence of the activated, phosphorylated form of JNK, whereas phospho-JNK expression is decreased by 50% to 60% in carotid thrombi in CD36-null mice,¹² similar to the decrease in MP mass in thrombi from CD36-null mice.

CONCLUSION

These experiments suggest that CD36 has both a proatherogenic and a prothrombotic role in the vascular system. Macrophage CD36 promotes foam cell formation and plaque formation. Platelet CD36 promotes thrombosis by signaling in response to oxLDL and by phospholipids present in cell-derived MP. Therefore, targeting CD36 or CD36-signaling pathways could be a strategy in the treatment of atheroinflammatory disorders and deserves exploration.

This is a momentous occasion for me, for the extraordinary people in the Clinical Neurocardiology Section at the National Institutes of Health (NIH), and for my family—my wife Minka and son Joey drove all the way from Maryland late last night and early this morning to be here. I thank them publicly here.

THE ‘SPARKLE OF INSIGHT’ FROM ENLIGHTENED INDUCTION

In these brief comments, as I look back on the road I have taken over the past 40 years carrying out patient-oriented research in heart-brain medicine, I would like to convey a viewpoint instead of dwelling on the presentation of research data.

The idea I wish to convey is that ignorance isn’t biased. If you have a hypothesis you want to test, you are inherently biased to find something positive—and, if you are in academic medicine, publishable—in the data you obtain. But if you have the technical capability to measure something no one else can measure, and you have sufficient mastery of the topic to know what is not yet known, then if you make an observation that you did not predict and if you recognize its significance, you have made a discovery. You have revealed a bit of the truth. You experience the highest joy and thrill a scientist can feel—a “sparkle of insight.” When this happens, if you have sense, you stop what you have been doing to pursue that discovery.

Hardly anyone has received a Nobel Prize for testing a theory, but many Nobel Prizes have been awarded for technological advances and for discoveries based on those advances. In my view, discoverers use an enlightened inductive approach at least as much as deduction. They develop new technology that enables key novel measurements, and they keep in mind gaps in knowledge, so that they are ready to appreciate the significance of their observations.

A PERSONAL EXAMPLE

‘You have to measure something’

Let me share an example of this process by relating a sparkle of insight I had several years ago. When I began working at the NIH, I met with the chief of the Hypertension-Endocrine Branch of the National Heart, Lung, and Blood Institute about the research program I would pursue. After listening patiently to me for many minutes as I spouted about how I was going to test hypotheses derived from the concepts that people with hypertension are “hyper-tense,” and that stress causes heart disease, the chief responded, “Well, these ideas are all well and good. But what are you going to measure? You can measure whatever you want, but you have to measure something.”

Measure something. I wanted to see if there was hyperactivity of the sympathetic nervous system or excessive sympathetic innervation in hypertension, and I started working on ways to measure sympathetic activity.

The sympathetic nervous system at a glance

First I should introduce you to the sympathetic nervous system, which is one of the main effectors by which the brain regulates the heart and blood vessels. It is a key link between the brain and heart. The sympathetic nerves to the heart and other organs do not come directly from the brain but from ganglia, which are clumps of nerve cell bodies strung like pearls on a necklace on each side of the spinal column. This origin outside the central nervous system will be an important fact to keep in mind.

In the heart, the sympathetic nerves travel with the coronary arteries and then dive into the heart muscle from the outside. Sympathetic nerves also enmesh the walls of arteries and arterioles. The arterioles constitute the main determinant of total peripheral resistance to blood flow in the body and therefore figure prominently in the control of blood pressure. The architectural association between sympathetic nerves and the muscle in the heart and arteriolar walls has enticed hypertension researchers for many decades.

A false start with plasma norepinephrine measurement

I developed novel methods for measuring plasma levels of norepinephrine, which is the chemical messenger that the sympathetic nervous system uses in regulation of the circulation, and of adrenaline (epinephrine), which is the well-known and potent “fight-or-flight” hormone.¹ Applying this technology to patients with high blood pressure led to several publications^2–9 but actually shed more heat than light on the hypothesis of sympathetic hyperactivity as a cause of or contributor to hypertension. In the face of negative data, the theory was qualified—sympathetic hyperactivity might be apparent only in the young, or the thin, or the Caucasian, or the male—but not abandoned.

Insights from visualizing sympathetic nerves in the heart

Then I embarked on a project to visualize sympathetic nerves in the heart, by a new technology called positron emission tomographic (PET) scanning. With several colleagues—including Irwin J. Kopin, Graeme Eisenhofer, Peter Chang, David Hovevey-Zion, Ehud Grossman, and Courtney Holmes—to whom I will always be grateful, I developed a PET imaging agent called 6-[¹⁸F]fluorodopamine.^10–13

Normally, PET scans using 6-[¹⁸F]fluorodopamine look remarkably similar to scans using ¹³N-labeled ammonia, a perfusion imaging agent. The first patient I studied with this new technology was a patient with a rare disease called pure autonomic failure (PAF). In PAF, there was already good evidence for a loss of sympathetic nerves throughout the body. Myocardial perfusion in this patient was normal, but there was much less than normal 6-[¹⁸F]fluorodopamine-derived radioactivity in the heart muscle. In another uncommon disease, multiple system atrophy (MSA), the perfusion was also normal, and the cardiac sympathetic nerves seemed intact, in line with what was already known about this progressive neurodegenerative disease.

Then I tested a patient who had been thought to have MSA but actually had Parkinson disease (PD) with orthostatic hypotension (a fall in blood pressure each time the person stands up). PD with orthostatic hypotension can be very difficult to distinguish from the parkinsonian form of MSA. To my complete surprise, the patient with PD had a remarkable decrease in 6-[¹⁸F]fluorodopa mine-derived radioactivity in the heart muscle. There was normal blood flow to the heart muscle, so the 6-[¹⁸F]fluorodopamine was being delivered, but there was no evidence of sympathetic nerves in the heart. The scans resembled those in the PAF patient, not the MSA patient.

This finding did not arise from a prediction to test a hypothesis. It wasn’t long before I tested additional PD patients and found the same unexpected results.^14,15 Because I was ignorant, I wasn’t biased. I felt I had put my finger on a piece of the truth, and I had to stop and think about the implications of this discovery. I never did come to test the hypotheses that I had sought out originally to test. Instead, I followed a totally new path, based on the discovery of cardiac sympathetic denervation in PD.

More than 50 neuroimaging studies since our original report have agreed remarkably consistently on the association between PD and loss of sympathetic nerves in the heart; moreover, postmortem pathology studies have amply confirmed that a profound loss of cardiac sympathetic nerves is characteristic of PD.¹⁶ I have yet to come across a single patient with PD and orthostatic hypotension who has not had cardiac sympathetic denervation, and virtually all patients with PD who do not have orthostatic hypotension seem to have at least partial loss of cardiac sympathetic nerves.

Considering that the source of those nerves is the ganglia, which lie outside the central nervous system, PD must be more than a brain disease and more than a movement disorder. It must also be a disease of the sympathetic nerves in the heart, a form of a dysautonomia, and a heart-brain disorder.

The role of catecholamines: Another discovery born of unbiased ignorance

Almost a half century ago, Hornykiewicz and colleagues made the pivotal discovery that PD features loss of dopamine in the nigrostriatal system in the brain.¹⁸ Given the cardiac sympathetic denervation, PD might be a disease of catecholamine systems both inside and outside the central nervous system—dopamine in the nigrostriatal system, and norepinephrine in the sympathetic nerves of the heart.

Then what of the third catecholamine, adrenaline, in PD? Plasma levels of adrenaline and of its metabolite, metanephrine, are normal in PD, even in patients who have PD and orthostatic hypotension, which involves loss of norepinephrine-producing nerves not only in the heart but in other organs.¹⁹ What is different about the adrenaline-producing cells in the medulla (from the Latin for “marrow”) of the adrenal glands atop each kidney? Why aren’t these catecholamine-producing cells also lost in PD?

I have some ideas in mind but won’t go into them here. The point is that the discovery of normal adrenaline-producing cells in PD, despite loss of cells producing the other catecholamines, was not based on my testing a hypothesis. It was a discovery born of ignorance, and because ignorance isn’t biased, that discovery points to the truth. Whatever the eventual explanation for the specific pattern of catecholamine cell loss in PD, it cannot refute the discovery itself.

HOW DISCOVERIES ARISE: AN APPLIED EXERCISE FOR READERS

PAF is a rare disease, and I have only studied several cases with high-resolution PET scanning of the brain, but so far they have all had this unexpected, unpredicted finding of loss of dopaminergic neurons in the substantia nigra.²⁰

What does this pattern mean? If PAF patients have just as much loss of nigral neurons as PD patients do, and if PAF patients do not have parkinsonism, then the movement disorder in PD cannot result from loss of the dopamine neurons in the substantia nigra per se. Instead, the movement disorder in PD seems to come from loss of the dopaminergic terminals in the striatum.

How can PAF patients have normal dopamine terminals in the putamen when the number of dopaminergic cell bodies is severely reduced? Somehow, PAF patients must be able to sprout new terminals, even as they lose the cell bodies. Maybe if we knew how PAF patients do this, we would have a way to treat or even prevent PD.

How do PAF patients maintain normal dopamine terminals as the cell bodies die off? No one knows. Until now, no one thought of asking such a question. No one hypothesized that this discovery would be made, but it was. And because ignorance isn’t biased, we have put our finger on the truth. By keeping in mind what isn’t known, we could see what wasn’t there. Now we can begin to think of what to look for next.

SUMMARY AND CONCLUSIONS

Because ignorance isn’t biased, if you have the tools to make relevant measurements, if you have sufficient mastery of the subject to know what isn’t known, and if you have access to patients with rare but informative disorders, you can make important discoveries based on inductions from observations.

The discoveries that cardiac sympathetic denervation characterizes PD and that parkinsonism does not result from loss of dopamine neurons per se depended crucially on studying patients with a rare disease, PAF. In 1657, William Harvey—the same William Harvey who first described the circulation of the blood and who first pointed out the effects of emotions on the heart—wrote eloquently about the extraordinary power of studying patients with rare diseases:Nature is nowhere accustomed more openly to display her secret mysteries than in cases where she shows traces of her workings apart from the beaten path; nor is there any better way to advance the proper practice of medicine than to give our minds to the discovery of the usual law of nature, by the careful investigation of cases of rarer forms of disease. For it has been found in almost all things, that what they contain of use or of application, is hardly perceived unless we are deprived of them, or they become deranged in some way.²¹

I hope I have convinced you of the importance of seeing what isn’t there. My thanks go out again to the Earl and Doris Bakken Heart-Brain Institute for this prestigious award, to my family, to my colleagues and friends, and to my patients. As I have written in Adrenaline and the Inner World: An Introduction to Scientific Integrative Medicine,¹⁷ patients serve as a unique scientific resource. They report what is wrong; they tell us the truth. We have to make sense of what they teach.

This is a momentous occasion for me, for the extraordinary people in the Clinical Neurocardiology Section at the National Institutes of Health (NIH), and for my family—my wife Minka and son Joey drove all the way from Maryland late last night and early this morning to be here. I thank them publicly here.

THE ‘SPARKLE OF INSIGHT’ FROM ENLIGHTENED INDUCTION

In these brief comments, as I look back on the road I have taken over the past 40 years carrying out patient-oriented research in heart-brain medicine, I would like to convey a viewpoint instead of dwelling on the presentation of research data.

The idea I wish to convey is that ignorance isn’t biased. If you have a hypothesis you want to test, you are inherently biased to find something positive—and, if you are in academic medicine, publishable—in the data you obtain. But if you have the technical capability to measure something no one else can measure, and you have sufficient mastery of the topic to know what is not yet known, then if you make an observation that you did not predict and if you recognize its significance, you have made a discovery. You have revealed a bit of the truth. You experience the highest joy and thrill a scientist can feel—a “sparkle of insight.” When this happens, if you have sense, you stop what you have been doing to pursue that discovery.

Hardly anyone has received a Nobel Prize for testing a theory, but many Nobel Prizes have been awarded for technological advances and for discoveries based on those advances. In my view, discoverers use an enlightened inductive approach at least as much as deduction. They develop new technology that enables key novel measurements, and they keep in mind gaps in knowledge, so that they are ready to appreciate the significance of their observations.

A PERSONAL EXAMPLE

‘You have to measure something’

Let me share an example of this process by relating a sparkle of insight I had several years ago. When I began working at the NIH, I met with the chief of the Hypertension-Endocrine Branch of the National Heart, Lung, and Blood Institute about the research program I would pursue. After listening patiently to me for many minutes as I spouted about how I was going to test hypotheses derived from the concepts that people with hypertension are “hyper-tense,” and that stress causes heart disease, the chief responded, “Well, these ideas are all well and good. But what are you going to measure? You can measure whatever you want, but you have to measure something.”

Measure something. I wanted to see if there was hyperactivity of the sympathetic nervous system or excessive sympathetic innervation in hypertension, and I started working on ways to measure sympathetic activity.

The sympathetic nervous system at a glance

First I should introduce you to the sympathetic nervous system, which is one of the main effectors by which the brain regulates the heart and blood vessels. It is a key link between the brain and heart. The sympathetic nerves to the heart and other organs do not come directly from the brain but from ganglia, which are clumps of nerve cell bodies strung like pearls on a necklace on each side of the spinal column. This origin outside the central nervous system will be an important fact to keep in mind.

In the heart, the sympathetic nerves travel with the coronary arteries and then dive into the heart muscle from the outside. Sympathetic nerves also enmesh the walls of arteries and arterioles. The arterioles constitute the main determinant of total peripheral resistance to blood flow in the body and therefore figure prominently in the control of blood pressure. The architectural association between sympathetic nerves and the muscle in the heart and arteriolar walls has enticed hypertension researchers for many decades.

A false start with plasma norepinephrine measurement

I developed novel methods for measuring plasma levels of norepinephrine, which is the chemical messenger that the sympathetic nervous system uses in regulation of the circulation, and of adrenaline (epinephrine), which is the well-known and potent “fight-or-flight” hormone.¹ Applying this technology to patients with high blood pressure led to several publications^2–9 but actually shed more heat than light on the hypothesis of sympathetic hyperactivity as a cause of or contributor to hypertension. In the face of negative data, the theory was qualified—sympathetic hyperactivity might be apparent only in the young, or the thin, or the Caucasian, or the male—but not abandoned.

Insights from visualizing sympathetic nerves in the heart

Then I embarked on a project to visualize sympathetic nerves in the heart, by a new technology called positron emission tomographic (PET) scanning. With several colleagues—including Irwin J. Kopin, Graeme Eisenhofer, Peter Chang, David Hovevey-Zion, Ehud Grossman, and Courtney Holmes—to whom I will always be grateful, I developed a PET imaging agent called 6-[¹⁸F]fluorodopamine.^10–13

Normally, PET scans using 6-[¹⁸F]fluorodopamine look remarkably similar to scans using ¹³N-labeled ammonia, a perfusion imaging agent. The first patient I studied with this new technology was a patient with a rare disease called pure autonomic failure (PAF). In PAF, there was already good evidence for a loss of sympathetic nerves throughout the body. Myocardial perfusion in this patient was normal, but there was much less than normal 6-[¹⁸F]fluorodopamine-derived radioactivity in the heart muscle. In another uncommon disease, multiple system atrophy (MSA), the perfusion was also normal, and the cardiac sympathetic nerves seemed intact, in line with what was already known about this progressive neurodegenerative disease.

Then I tested a patient who had been thought to have MSA but actually had Parkinson disease (PD) with orthostatic hypotension (a fall in blood pressure each time the person stands up). PD with orthostatic hypotension can be very difficult to distinguish from the parkinsonian form of MSA. To my complete surprise, the patient with PD had a remarkable decrease in 6-[¹⁸F]fluorodopa mine-derived radioactivity in the heart muscle. There was normal blood flow to the heart muscle, so the 6-[¹⁸F]fluorodopamine was being delivered, but there was no evidence of sympathetic nerves in the heart. The scans resembled those in the PAF patient, not the MSA patient.

This finding did not arise from a prediction to test a hypothesis. It wasn’t long before I tested additional PD patients and found the same unexpected results.^14,15 Because I was ignorant, I wasn’t biased. I felt I had put my finger on a piece of the truth, and I had to stop and think about the implications of this discovery. I never did come to test the hypotheses that I had sought out originally to test. Instead, I followed a totally new path, based on the discovery of cardiac sympathetic denervation in PD.

More than 50 neuroimaging studies since our original report have agreed remarkably consistently on the association between PD and loss of sympathetic nerves in the heart; moreover, postmortem pathology studies have amply confirmed that a profound loss of cardiac sympathetic nerves is characteristic of PD.¹⁶ I have yet to come across a single patient with PD and orthostatic hypotension who has not had cardiac sympathetic denervation, and virtually all patients with PD who do not have orthostatic hypotension seem to have at least partial loss of cardiac sympathetic nerves.

Considering that the source of those nerves is the ganglia, which lie outside the central nervous system, PD must be more than a brain disease and more than a movement disorder. It must also be a disease of the sympathetic nerves in the heart, a form of a dysautonomia, and a heart-brain disorder.

The role of catecholamines: Another discovery born of unbiased ignorance

Almost a half century ago, Hornykiewicz and colleagues made the pivotal discovery that PD features loss of dopamine in the nigrostriatal system in the brain.¹⁸ Given the cardiac sympathetic denervation, PD might be a disease of catecholamine systems both inside and outside the central nervous system—dopamine in the nigrostriatal system, and norepinephrine in the sympathetic nerves of the heart.

Then what of the third catecholamine, adrenaline, in PD? Plasma levels of adrenaline and of its metabolite, metanephrine, are normal in PD, even in patients who have PD and orthostatic hypotension, which involves loss of norepinephrine-producing nerves not only in the heart but in other organs.¹⁹ What is different about the adrenaline-producing cells in the medulla (from the Latin for “marrow”) of the adrenal glands atop each kidney? Why aren’t these catecholamine-producing cells also lost in PD?

I have some ideas in mind but won’t go into them here. The point is that the discovery of normal adrenaline-producing cells in PD, despite loss of cells producing the other catecholamines, was not based on my testing a hypothesis. It was a discovery born of ignorance, and because ignorance isn’t biased, that discovery points to the truth. Whatever the eventual explanation for the specific pattern of catecholamine cell loss in PD, it cannot refute the discovery itself.

HOW DISCOVERIES ARISE: AN APPLIED EXERCISE FOR READERS

PAF is a rare disease, and I have only studied several cases with high-resolution PET scanning of the brain, but so far they have all had this unexpected, unpredicted finding of loss of dopaminergic neurons in the substantia nigra.²⁰

What does this pattern mean? If PAF patients have just as much loss of nigral neurons as PD patients do, and if PAF patients do not have parkinsonism, then the movement disorder in PD cannot result from loss of the dopamine neurons in the substantia nigra per se. Instead, the movement disorder in PD seems to come from loss of the dopaminergic terminals in the striatum.

How can PAF patients have normal dopamine terminals in the putamen when the number of dopaminergic cell bodies is severely reduced? Somehow, PAF patients must be able to sprout new terminals, even as they lose the cell bodies. Maybe if we knew how PAF patients do this, we would have a way to treat or even prevent PD.

How do PAF patients maintain normal dopamine terminals as the cell bodies die off? No one knows. Until now, no one thought of asking such a question. No one hypothesized that this discovery would be made, but it was. And because ignorance isn’t biased, we have put our finger on the truth. By keeping in mind what isn’t known, we could see what wasn’t there. Now we can begin to think of what to look for next.

SUMMARY AND CONCLUSIONS

Because ignorance isn’t biased, if you have the tools to make relevant measurements, if you have sufficient mastery of the subject to know what isn’t known, and if you have access to patients with rare but informative disorders, you can make important discoveries based on inductions from observations.

The discoveries that cardiac sympathetic denervation characterizes PD and that parkinsonism does not result from loss of dopamine neurons per se depended crucially on studying patients with a rare disease, PAF. In 1657, William Harvey—the same William Harvey who first described the circulation of the blood and who first pointed out the effects of emotions on the heart—wrote eloquently about the extraordinary power of studying patients with rare diseases:Nature is nowhere accustomed more openly to display her secret mysteries than in cases where she shows traces of her workings apart from the beaten path; nor is there any better way to advance the proper practice of medicine than to give our minds to the discovery of the usual law of nature, by the careful investigation of cases of rarer forms of disease. For it has been found in almost all things, that what they contain of use or of application, is hardly perceived unless we are deprived of them, or they become deranged in some way.²¹

I hope I have convinced you of the importance of seeing what isn’t there. My thanks go out again to the Earl and Doris Bakken Heart-Brain Institute for this prestigious award, to my family, to my colleagues and friends, and to my patients. As I have written in Adrenaline and the Inner World: An Introduction to Scientific Integrative Medicine,¹⁷ patients serve as a unique scientific resource. They report what is wrong; they tell us the truth. We have to make sense of what they teach.

This is a momentous occasion for me, for the extraordinary people in the Clinical Neurocardiology Section at the National Institutes of Health (NIH), and for my family—my wife Minka and son Joey drove all the way from Maryland late last night and early this morning to be here. I thank them publicly here.

THE ‘SPARKLE OF INSIGHT’ FROM ENLIGHTENED INDUCTION

In these brief comments, as I look back on the road I have taken over the past 40 years carrying out patient-oriented research in heart-brain medicine, I would like to convey a viewpoint instead of dwelling on the presentation of research data.

The idea I wish to convey is that ignorance isn’t biased. If you have a hypothesis you want to test, you are inherently biased to find something positive—and, if you are in academic medicine, publishable—in the data you obtain. But if you have the technical capability to measure something no one else can measure, and you have sufficient mastery of the topic to know what is not yet known, then if you make an observation that you did not predict and if you recognize its significance, you have made a discovery. You have revealed a bit of the truth. You experience the highest joy and thrill a scientist can feel—a “sparkle of insight.” When this happens, if you have sense, you stop what you have been doing to pursue that discovery.

Hardly anyone has received a Nobel Prize for testing a theory, but many Nobel Prizes have been awarded for technological advances and for discoveries based on those advances. In my view, discoverers use an enlightened inductive approach at least as much as deduction. They develop new technology that enables key novel measurements, and they keep in mind gaps in knowledge, so that they are ready to appreciate the significance of their observations.

A PERSONAL EXAMPLE

‘You have to measure something’

Let me share an example of this process by relating a sparkle of insight I had several years ago. When I began working at the NIH, I met with the chief of the Hypertension-Endocrine Branch of the National Heart, Lung, and Blood Institute about the research program I would pursue. After listening patiently to me for many minutes as I spouted about how I was going to test hypotheses derived from the concepts that people with hypertension are “hyper-tense,” and that stress causes heart disease, the chief responded, “Well, these ideas are all well and good. But what are you going to measure? You can measure whatever you want, but you have to measure something.”

Measure something. I wanted to see if there was hyperactivity of the sympathetic nervous system or excessive sympathetic innervation in hypertension, and I started working on ways to measure sympathetic activity.

The sympathetic nervous system at a glance

First I should introduce you to the sympathetic nervous system, which is one of the main effectors by which the brain regulates the heart and blood vessels. It is a key link between the brain and heart. The sympathetic nerves to the heart and other organs do not come directly from the brain but from ganglia, which are clumps of nerve cell bodies strung like pearls on a necklace on each side of the spinal column. This origin outside the central nervous system will be an important fact to keep in mind.

In the heart, the sympathetic nerves travel with the coronary arteries and then dive into the heart muscle from the outside. Sympathetic nerves also enmesh the walls of arteries and arterioles. The arterioles constitute the main determinant of total peripheral resistance to blood flow in the body and therefore figure prominently in the control of blood pressure. The architectural association between sympathetic nerves and the muscle in the heart and arteriolar walls has enticed hypertension researchers for many decades.

A false start with plasma norepinephrine measurement

I developed novel methods for measuring plasma levels of norepinephrine, which is the chemical messenger that the sympathetic nervous system uses in regulation of the circulation, and of adrenaline (epinephrine), which is the well-known and potent “fight-or-flight” hormone.¹ Applying this technology to patients with high blood pressure led to several publications^2–9 but actually shed more heat than light on the hypothesis of sympathetic hyperactivity as a cause of or contributor to hypertension. In the face of negative data, the theory was qualified—sympathetic hyperactivity might be apparent only in the young, or the thin, or the Caucasian, or the male—but not abandoned.

Insights from visualizing sympathetic nerves in the heart

Then I embarked on a project to visualize sympathetic nerves in the heart, by a new technology called positron emission tomographic (PET) scanning. With several colleagues—including Irwin J. Kopin, Graeme Eisenhofer, Peter Chang, David Hovevey-Zion, Ehud Grossman, and Courtney Holmes—to whom I will always be grateful, I developed a PET imaging agent called 6-[¹⁸F]fluorodopamine.^10–13

Normally, PET scans using 6-[¹⁸F]fluorodopamine look remarkably similar to scans using ¹³N-labeled ammonia, a perfusion imaging agent. The first patient I studied with this new technology was a patient with a rare disease called pure autonomic failure (PAF). In PAF, there was already good evidence for a loss of sympathetic nerves throughout the body. Myocardial perfusion in this patient was normal, but there was much less than normal 6-[¹⁸F]fluorodopamine-derived radioactivity in the heart muscle. In another uncommon disease, multiple system atrophy (MSA), the perfusion was also normal, and the cardiac sympathetic nerves seemed intact, in line with what was already known about this progressive neurodegenerative disease.

Then I tested a patient who had been thought to have MSA but actually had Parkinson disease (PD) with orthostatic hypotension (a fall in blood pressure each time the person stands up). PD with orthostatic hypotension can be very difficult to distinguish from the parkinsonian form of MSA. To my complete surprise, the patient with PD had a remarkable decrease in 6-[¹⁸F]fluorodopa mine-derived radioactivity in the heart muscle. There was normal blood flow to the heart muscle, so the 6-[¹⁸F]fluorodopamine was being delivered, but there was no evidence of sympathetic nerves in the heart. The scans resembled those in the PAF patient, not the MSA patient.

This finding did not arise from a prediction to test a hypothesis. It wasn’t long before I tested additional PD patients and found the same unexpected results.^14,15 Because I was ignorant, I wasn’t biased. I felt I had put my finger on a piece of the truth, and I had to stop and think about the implications of this discovery. I never did come to test the hypotheses that I had sought out originally to test. Instead, I followed a totally new path, based on the discovery of cardiac sympathetic denervation in PD.

More than 50 neuroimaging studies since our original report have agreed remarkably consistently on the association between PD and loss of sympathetic nerves in the heart; moreover, postmortem pathology studies have amply confirmed that a profound loss of cardiac sympathetic nerves is characteristic of PD.¹⁶ I have yet to come across a single patient with PD and orthostatic hypotension who has not had cardiac sympathetic denervation, and virtually all patients with PD who do not have orthostatic hypotension seem to have at least partial loss of cardiac sympathetic nerves.

Considering that the source of those nerves is the ganglia, which lie outside the central nervous system, PD must be more than a brain disease and more than a movement disorder. It must also be a disease of the sympathetic nerves in the heart, a form of a dysautonomia, and a heart-brain disorder.

The role of catecholamines: Another discovery born of unbiased ignorance

Almost a half century ago, Hornykiewicz and colleagues made the pivotal discovery that PD features loss of dopamine in the nigrostriatal system in the brain.¹⁸ Given the cardiac sympathetic denervation, PD might be a disease of catecholamine systems both inside and outside the central nervous system—dopamine in the nigrostriatal system, and norepinephrine in the sympathetic nerves of the heart.

Then what of the third catecholamine, adrenaline, in PD? Plasma levels of adrenaline and of its metabolite, metanephrine, are normal in PD, even in patients who have PD and orthostatic hypotension, which involves loss of norepinephrine-producing nerves not only in the heart but in other organs.¹⁹ What is different about the adrenaline-producing cells in the medulla (from the Latin for “marrow”) of the adrenal glands atop each kidney? Why aren’t these catecholamine-producing cells also lost in PD?

I have some ideas in mind but won’t go into them here. The point is that the discovery of normal adrenaline-producing cells in PD, despite loss of cells producing the other catecholamines, was not based on my testing a hypothesis. It was a discovery born of ignorance, and because ignorance isn’t biased, that discovery points to the truth. Whatever the eventual explanation for the specific pattern of catecholamine cell loss in PD, it cannot refute the discovery itself.

HOW DISCOVERIES ARISE: AN APPLIED EXERCISE FOR READERS

PAF is a rare disease, and I have only studied several cases with high-resolution PET scanning of the brain, but so far they have all had this unexpected, unpredicted finding of loss of dopaminergic neurons in the substantia nigra.²⁰

What does this pattern mean? If PAF patients have just as much loss of nigral neurons as PD patients do, and if PAF patients do not have parkinsonism, then the movement disorder in PD cannot result from loss of the dopamine neurons in the substantia nigra per se. Instead, the movement disorder in PD seems to come from loss of the dopaminergic terminals in the striatum.

How can PAF patients have normal dopamine terminals in the putamen when the number of dopaminergic cell bodies is severely reduced? Somehow, PAF patients must be able to sprout new terminals, even as they lose the cell bodies. Maybe if we knew how PAF patients do this, we would have a way to treat or even prevent PD.

How do PAF patients maintain normal dopamine terminals as the cell bodies die off? No one knows. Until now, no one thought of asking such a question. No one hypothesized that this discovery would be made, but it was. And because ignorance isn’t biased, we have put our finger on the truth. By keeping in mind what isn’t known, we could see what wasn’t there. Now we can begin to think of what to look for next.

SUMMARY AND CONCLUSIONS

Because ignorance isn’t biased, if you have the tools to make relevant measurements, if you have sufficient mastery of the subject to know what isn’t known, and if you have access to patients with rare but informative disorders, you can make important discoveries based on inductions from observations.

The discoveries that cardiac sympathetic denervation characterizes PD and that parkinsonism does not result from loss of dopamine neurons per se depended crucially on studying patients with a rare disease, PAF. In 1657, William Harvey—the same William Harvey who first described the circulation of the blood and who first pointed out the effects of emotions on the heart—wrote eloquently about the extraordinary power of studying patients with rare diseases:Nature is nowhere accustomed more openly to display her secret mysteries than in cases where she shows traces of her workings apart from the beaten path; nor is there any better way to advance the proper practice of medicine than to give our minds to the discovery of the usual law of nature, by the careful investigation of cases of rarer forms of disease. For it has been found in almost all things, that what they contain of use or of application, is hardly perceived unless we are deprived of them, or they become deranged in some way.²¹

I hope I have convinced you of the importance of seeing what isn’t there. My thanks go out again to the Earl and Doris Bakken Heart-Brain Institute for this prestigious award, to my family, to my colleagues and friends, and to my patients. As I have written in Adrenaline and the Inner World: An Introduction to Scientific Integrative Medicine,¹⁷ patients serve as a unique scientific resource. They report what is wrong; they tell us the truth. We have to make sense of what they teach.

Heart rate variability (HRV) has been a focus of interest in cardiovascular physiology for more than 150 years. This review will briefly survey the history of research linking HRV to respiration and then explore the clinical significance of this linkage, with a focus on HRV with deep breathing.

HRV AND RESPIRATION: THE EARLY RESEARCH

The first report linking HRV to respiration has been credited to Karl Ludwig, who in 1847 noted that heart rate increased with inspiration and decreased with expiration.^1,2 The precise origin of this variability has been studied extensively, but a single unifying mechanism defining the determinants of HRV with respiration has not been established. However, several mechanisms have been identified that may be contributing to HRV. Hering in 1871 noted in dog experiments that inflation of the lungs was associated with a tachycardia and that additional higher-pressure insufflation resulted in a bradycardia. He concluded that HRV was determined by pulmonary reflexes.^2,3 Bainbridge observed in dog experiments in 1915 that the heart rate increased during the diastolic filling of the heart that occurred during inspiration.⁴ In a subsequent article, published in 1920, Bainbridge attributed HRV to this reflex, which now carries his name.⁵

There is also evidence that HRV may be caused by central nervous system mechanisms. Canine experiments have revealed that rhythmic variations in the heart rate and ventricular pressure waves may coincide with rib cage movements in innervated, isovolumetric, left ventricular preparations.⁶ These data are consistent with radiation of respiratory center activity to the cardiovascular autonomic centers in the medulla resulting in HRV. There is also evidence that stretch of the right atrium and sinus node region may produce HRV via cardiac reflexes.⁷ It is likely that all of these mechanisms are contributing at some level to the HRV that is observed with respiration.

INSIGHTS INTO CLINICAL IMPLICATIONS

Reprinted from British Medical Journal (Wheeler T, Watkins PJ. Cardiac denervation in diabetes. Br Med J 1973;4:584–586) with permission from the BMJ Publishing Group.

Reprinted, with permission, from Journal of Applied Physiology (Katona PG, Jih F. Respiratory sinus arrhythmia: noninvasive measure of parasympathetic cardiac control. J Appl Physiol 1975;39:801–805).

Reprinted, with permission, from Journal of Applied Physiology (Fouad et al. Assessment of parasympathetic control of heart rate by a noninvasive method. Am J Physiol 1984; 246:H838–H842).

METHODS OF MEASURING HRV

A wide variety of methods have been developed to measure HRV.^11,12 Some of the methods employ statistical analysis, typically of prolonged recordings of 24 hours or longer. These methods include simple statistics such as the standard deviation of the heart rate or the R-R interval as well as more complex statistical measures such as the mean squared successive difference of the R-R intervals. These methods have been applied mostly to the analysis of prognosis following acute myocardial infarction. Reduced HRV has been established as a powerful predictor of mortality and arrhythmic complications following acute myocardial infarction.¹¹ The methods developed for clinical tests of cardiovagal function typically involve measuring HRVdb over short intervals (< 90 sec). Deep breathing magnifies HRV with respiration, allowing for methods to assess HRV with respiratory cycles.

FACTORS THAT AFFECT HRV WITH DEEP BREATHING

Many variables may affect HRVdb.¹² HRVdb is influenced by age, as the variability decreases with advancing age, so it is essential to use methods with well-defined age-stratified normal values. HRVdb is maximal when the patient is lying supine and breathing at a rate of 5 to 6 breaths per minute. The depth of breathing for a maximum result requires a tidal volume of approximately 1.2 L for an average adult. Protocols that involve breathing for more than 90 seconds may induce hypocapnea, which can reduce HRVdb. Most importantly, numerous medications can affect HRVdb. Medications with anticholinergic activity, including over-the-counter cold medications, tricylic antidepressants, and antispasmodics, should be discontinued at least 48 hours prior to testing, if possible. Patients are also instructed to not drink caffeinated beverages, use nicotine, or drink alcohol 3 hours prior to testing.

CLINICAL APPLICATIONS

HRVdb represents a very sensitive measure of cardiovagal or parasympathetic cardiac function and thus is an important component of the battery of cardiovascular autonomic function tests used in clinical autonomic laboratories. In most autonomic disorders, parasympathetic function is affected before sympathetic function, so HRVdb provides a sensitive screening measure for parasympathetic dysfunction in many autonomic disorders. HRVdb has proven to be a sensitive and reliable clinical test for the early detection of cardiovagal dysfunction in a wide spectrum of autonomic disorders, including diabetic autonomic neuropathy,¹⁴ uremic neuropathy,¹⁷ familial autonomic neuropathies,¹⁸ and various small fiber neuropathies.^19,20 HRVdb has also been valuable in assessing patients with pure autonomic failure,²¹ multisystem atrophy,²² and other central neurodegenerative disorders.²³

Heart rate variability (HRV) has been a focus of interest in cardiovascular physiology for more than 150 years. This review will briefly survey the history of research linking HRV to respiration and then explore the clinical significance of this linkage, with a focus on HRV with deep breathing.

HRV AND RESPIRATION: THE EARLY RESEARCH

The first report linking HRV to respiration has been credited to Karl Ludwig, who in 1847 noted that heart rate increased with inspiration and decreased with expiration.^1,2 The precise origin of this variability has been studied extensively, but a single unifying mechanism defining the determinants of HRV with respiration has not been established. However, several mechanisms have been identified that may be contributing to HRV. Hering in 1871 noted in dog experiments that inflation of the lungs was associated with a tachycardia and that additional higher-pressure insufflation resulted in a bradycardia. He concluded that HRV was determined by pulmonary reflexes.^2,3 Bainbridge observed in dog experiments in 1915 that the heart rate increased during the diastolic filling of the heart that occurred during inspiration.⁴ In a subsequent article, published in 1920, Bainbridge attributed HRV to this reflex, which now carries his name.⁵

There is also evidence that HRV may be caused by central nervous system mechanisms. Canine experiments have revealed that rhythmic variations in the heart rate and ventricular pressure waves may coincide with rib cage movements in innervated, isovolumetric, left ventricular preparations.⁶ These data are consistent with radiation of respiratory center activity to the cardiovascular autonomic centers in the medulla resulting in HRV. There is also evidence that stretch of the right atrium and sinus node region may produce HRV via cardiac reflexes.⁷ It is likely that all of these mechanisms are contributing at some level to the HRV that is observed with respiration.

INSIGHTS INTO CLINICAL IMPLICATIONS

Reprinted from British Medical Journal (Wheeler T, Watkins PJ. Cardiac denervation in diabetes. Br Med J 1973;4:584–586) with permission from the BMJ Publishing Group.

Reprinted, with permission, from Journal of Applied Physiology (Katona PG, Jih F. Respiratory sinus arrhythmia: noninvasive measure of parasympathetic cardiac control. J Appl Physiol 1975;39:801–805).

Reprinted, with permission, from Journal of Applied Physiology (Fouad et al. Assessment of parasympathetic control of heart rate by a noninvasive method. Am J Physiol 1984; 246:H838–H842).

METHODS OF MEASURING HRV

A wide variety of methods have been developed to measure HRV.^11,12 Some of the methods employ statistical analysis, typically of prolonged recordings of 24 hours or longer. These methods include simple statistics such as the standard deviation of the heart rate or the R-R interval as well as more complex statistical measures such as the mean squared successive difference of the R-R intervals. These methods have been applied mostly to the analysis of prognosis following acute myocardial infarction. Reduced HRV has been established as a powerful predictor of mortality and arrhythmic complications following acute myocardial infarction.¹¹ The methods developed for clinical tests of cardiovagal function typically involve measuring HRVdb over short intervals (< 90 sec). Deep breathing magnifies HRV with respiration, allowing for methods to assess HRV with respiratory cycles.

FACTORS THAT AFFECT HRV WITH DEEP BREATHING

Many variables may affect HRVdb.¹² HRVdb is influenced by age, as the variability decreases with advancing age, so it is essential to use methods with well-defined age-stratified normal values. HRVdb is maximal when the patient is lying supine and breathing at a rate of 5 to 6 breaths per minute. The depth of breathing for a maximum result requires a tidal volume of approximately 1.2 L for an average adult. Protocols that involve breathing for more than 90 seconds may induce hypocapnea, which can reduce HRVdb. Most importantly, numerous medications can affect HRVdb. Medications with anticholinergic activity, including over-the-counter cold medications, tricylic antidepressants, and antispasmodics, should be discontinued at least 48 hours prior to testing, if possible. Patients are also instructed to not drink caffeinated beverages, use nicotine, or drink alcohol 3 hours prior to testing.

CLINICAL APPLICATIONS

HRVdb represents a very sensitive measure of cardiovagal or parasympathetic cardiac function and thus is an important component of the battery of cardiovascular autonomic function tests used in clinical autonomic laboratories. In most autonomic disorders, parasympathetic function is affected before sympathetic function, so HRVdb provides a sensitive screening measure for parasympathetic dysfunction in many autonomic disorders. HRVdb has proven to be a sensitive and reliable clinical test for the early detection of cardiovagal dysfunction in a wide spectrum of autonomic disorders, including diabetic autonomic neuropathy,¹⁴ uremic neuropathy,¹⁷ familial autonomic neuropathies,¹⁸ and various small fiber neuropathies.^19,20 HRVdb has also been valuable in assessing patients with pure autonomic failure,²¹ multisystem atrophy,²² and other central neurodegenerative disorders.²³

Heart rate variability (HRV) has been a focus of interest in cardiovascular physiology for more than 150 years. This review will briefly survey the history of research linking HRV to respiration and then explore the clinical significance of this linkage, with a focus on HRV with deep breathing.

HRV AND RESPIRATION: THE EARLY RESEARCH

The first report linking HRV to respiration has been credited to Karl Ludwig, who in 1847 noted that heart rate increased with inspiration and decreased with expiration.^1,2 The precise origin of this variability has been studied extensively, but a single unifying mechanism defining the determinants of HRV with respiration has not been established. However, several mechanisms have been identified that may be contributing to HRV. Hering in 1871 noted in dog experiments that inflation of the lungs was associated with a tachycardia and that additional higher-pressure insufflation resulted in a bradycardia. He concluded that HRV was determined by pulmonary reflexes.^2,3 Bainbridge observed in dog experiments in 1915 that the heart rate increased during the diastolic filling of the heart that occurred during inspiration.⁴ In a subsequent article, published in 1920, Bainbridge attributed HRV to this reflex, which now carries his name.⁵

There is also evidence that HRV may be caused by central nervous system mechanisms. Canine experiments have revealed that rhythmic variations in the heart rate and ventricular pressure waves may coincide with rib cage movements in innervated, isovolumetric, left ventricular preparations.⁶ These data are consistent with radiation of respiratory center activity to the cardiovascular autonomic centers in the medulla resulting in HRV. There is also evidence that stretch of the right atrium and sinus node region may produce HRV via cardiac reflexes.⁷ It is likely that all of these mechanisms are contributing at some level to the HRV that is observed with respiration.

INSIGHTS INTO CLINICAL IMPLICATIONS

Reprinted from British Medical Journal (Wheeler T, Watkins PJ. Cardiac denervation in diabetes. Br Med J 1973;4:584–586) with permission from the BMJ Publishing Group.

Reprinted, with permission, from Journal of Applied Physiology (Katona PG, Jih F. Respiratory sinus arrhythmia: noninvasive measure of parasympathetic cardiac control. J Appl Physiol 1975;39:801–805).

Reprinted, with permission, from Journal of Applied Physiology (Fouad et al. Assessment of parasympathetic control of heart rate by a noninvasive method. Am J Physiol 1984; 246:H838–H842).

METHODS OF MEASURING HRV

A wide variety of methods have been developed to measure HRV.^11,12 Some of the methods employ statistical analysis, typically of prolonged recordings of 24 hours or longer. These methods include simple statistics such as the standard deviation of the heart rate or the R-R interval as well as more complex statistical measures such as the mean squared successive difference of the R-R intervals. These methods have been applied mostly to the analysis of prognosis following acute myocardial infarction. Reduced HRV has been established as a powerful predictor of mortality and arrhythmic complications following acute myocardial infarction.¹¹ The methods developed for clinical tests of cardiovagal function typically involve measuring HRVdb over short intervals (< 90 sec). Deep breathing magnifies HRV with respiration, allowing for methods to assess HRV with respiratory cycles.

FACTORS THAT AFFECT HRV WITH DEEP BREATHING

Many variables may affect HRVdb.¹² HRVdb is influenced by age, as the variability decreases with advancing age, so it is essential to use methods with well-defined age-stratified normal values. HRVdb is maximal when the patient is lying supine and breathing at a rate of 5 to 6 breaths per minute. The depth of breathing for a maximum result requires a tidal volume of approximately 1.2 L for an average adult. Protocols that involve breathing for more than 90 seconds may induce hypocapnea, which can reduce HRVdb. Most importantly, numerous medications can affect HRVdb. Medications with anticholinergic activity, including over-the-counter cold medications, tricylic antidepressants, and antispasmodics, should be discontinued at least 48 hours prior to testing, if possible. Patients are also instructed to not drink caffeinated beverages, use nicotine, or drink alcohol 3 hours prior to testing.

CLINICAL APPLICATIONS

HRVdb represents a very sensitive measure of cardiovagal or parasympathetic cardiac function and thus is an important component of the battery of cardiovascular autonomic function tests used in clinical autonomic laboratories. In most autonomic disorders, parasympathetic function is affected before sympathetic function, so HRVdb provides a sensitive screening measure for parasympathetic dysfunction in many autonomic disorders. HRVdb has proven to be a sensitive and reliable clinical test for the early detection of cardiovagal dysfunction in a wide spectrum of autonomic disorders, including diabetic autonomic neuropathy,¹⁴ uremic neuropathy,¹⁷ familial autonomic neuropathies,¹⁸ and various small fiber neuropathies.^19,20 HRVdb has also been valuable in assessing patients with pure autonomic failure,²¹ multisystem atrophy,²² and other central neurodegenerative disorders.²³

RATIONALE

In the industrialized world, average life expectancy has nearly doubled since the 19th century. One of the consequences of this increase in life span is that the sequelae of diseases also have increased. For coronary artery disease (CAD), one of the most prevalent diseases in the western world, this has resulted in an amplification of the number of patients suffering from heart failure, arrhythmias, and refractory angina. Much progress has recently been made in nonpharmacologic therapies for these deleterious consequences of CAD, such as cardiac resynchronization for heart failure, implantable defibrillators for ventricular arrhythmias, and electrical neuromodulation by means of spinal cord stimulation for chronic angina that is refractory to conventional strategies.

For patients suffering from severe angina secondary to end-stage CAD who have no other options to alleviate their complaints, electrical neuromodulation may be the preferred adjunctive treatment.¹ Although spinal cord stimulation is still not approved by the US Food and Drug Administration for treatment of refractory angina, it is is accepted in the American College of Cardiology/American Heart Association guidelines for chronic stable angina, with a class II indication, and is frequently used for this indication in Europe.²

However, to understand underlying mechanisms of therapies such as electrical neuromodulation—executed through either transcutaneous electrical nerve stimulation or spinal cord stimulation—for angina pectoris and to improve the effect and safety of these therapies, clinical questions concerning neuromodulation must be evaluated in experimental models. The outcomes of these preclinical experimental studies subsequently need to be assessed in humans.

Although therapeutic improvements from implantable devices would not have been possible without experimental work, any experimentation must be avoided if it is not approved by the relevant ethics committee(s) or is not conducted in keeping with standard guidelines. For this reason it is sometimes more feasible, when appropriate, to make use of simulation models—for instance, to study regularization of atrial fibrillation by means of a device.^3,4

PROJECT 1: EMOTIONS AND MYOCARDIAL ISCHEMIA

In 1772, Heberden described to physicians in England the clinical symptoms of exercise-induced chest discomfort, with its emotional component and vaguely distributed projection on the chest, as follows: “The seat of it, and sense of strangling and anxiety with which it is attended, may make it not improperly be called angina pectoris.”⁵ Since then, it has been demonstrated repeatedly that strong emotional distress frequently precedes or is associated with complaints of pain in the chest. Further, emotional suffering has been associated with increased mortality in patients with CAD. We and others, unfortunately, were confronted with very limited knowledge of the precise locations of the origin of emotions in the limbic structures of the forebrain. Even less was known about the relationship of these brain structures and the heart, owing to technical limitations in the field of neuroanatomical tract tracing, among other reasons. As a result, the nervous pathways from the heart, through which signals are propagated to the brain in order to activate emotional components, were not accurately identified. We therefore initiated Project 1 to study, in a rat model, neuroanatomical characterization of the neuronal circuitry controlling cardiac activity, specifically during cardiac distress.

In the area of identifying efferent neural pathways from the heart, we were the first to publish an experimental setup making use of a neurotropic herpesvirus from the Bartha strain of the pseudorabies virus (PRV).⁶ Following injections of PRV into the left and right myocardium or into atrial tissue, PRV infects the neurons that innervate the injection site and is then transported in the neural network, where the virus may cross at least four synapses. This transneuronal retrograde viral pathway labeling method with PRV provided us the opportunity to study cardiovascular controlling networks. The distribution of the PRV-infected cells was studied immunocytochemically after survival times of 3 to 6 days. Right ventricular infection showed labeling in the same nuclei as left ventricular labeling, but the number of PRV-positive cells was always higher and the localization of PRV within the nuclei differed. These obvious signs for differentiation within the nuclei suggest differential neuronal pathways to various parts of the heart.

Following injection of PRV at different cardiac sites, differences in density and localization of PRV-positive cells were found predominantly in higher-order neurons that are known to be involved in cardiac control. Transection of the spinal cord at Th1, performed to reveal selectively the parasympathetic neuronal networks, reduced the number of labeled cells, specifically in the periaqueductal gray matter. Virus-labeled sympathetic preganglionic cells were found in the Th1–Th7 thoracic intermediolateral cell groups, with some additional infections at Th8–Th11 after inoculations of the ventricular myocardium. The rostral parts of the insular cortex appeared to be linked selectively to sympathetic innervation of the heart.⁶

From the experiments we hypothesized that, according to the type of lesion, the pattern of cardiac innervation may account for a specific malfunctioning. Subsequently, the subendocardial clustered parasympathetic nerves make these nerves more vulnerable for myocardial damage than the superficial spread of sympathetic nerves. In this respect, the identification of three preganglionic parasympathetic nuclei in cardiac control—ie, the dorsal motor nucleus of the vagus (20% labeling), the nucleus ambiguus, and the periambiguus—constituted the most striking findings.

PROJECTS 2 AND 3: CARDIAC NOCICEPTOR ACTIVATION

The cortical structures and their related output pathways also serve as effector systems for initiation of autonomic and behavioral responses by forebrain neuronal networks that make us aware of cardiac pain. However, these cortical and subcortical structures involved in cardiac pain perception were more or less terra incognita. In addition, we studied fundamental aspects of cardiac nociceptor activation (Project 2) and transduction of cardiac pain (Project 3). Unfortunately, there was no experimental animal model for angina pectoris. The aim of these projects was to obtain, both in patients and in animals, knowledge about cardiac nociceptor activation mechanisms, the transmission and perception of cardiac pain, and behavioral and autonomic responses.

To enable the study of mechanisms of neurostimulation during episodes of acute cardiac pain, we worked out an animal model for angina pectoris. For that reason we experimented with models in which we created an acute myocardial infarction. We had to reject this model since surgery and, more importantly, anesthesia interfered with the patterns of cerebral expression of immediate early genes (c-fos, c-jun) triggered by cardiac pain and/or neurostimulation. However, a spinoff from this project was the observation that cardiac tissue damage causes a reproducible and selective cerebral endothelial leakage of immunoglobulin G (IgG) molecules. Follow-up experiments showed that proinflammatory cytokines, which are released into the circulation after cardiac tissue damage, can generate the same pattern of blood-brain barrier dysfunction⁷ (see Project 4).

We then experimented with infusions of capsaicin into the pericardial space of unrestrained and unanesthetized rats to induce acute cardiac pain. This model appeared to be very promising and allows visualization of the behavioral and autonomic responses to cardiac pain. Cerebral c-fos expression patterns, a marker for structures involved in cardiac pain transmission and perception, were studied and validated with positron emission tomography (PET) imaging in patients.⁸

Project 2: Nociception of cardiac pain in patients

To study relationships between neurotransmitters and other molecules that contribute to pain and psychological variables, we studied cardiac tissues obtained from 22 patients with angina during coronary artery bypass graft surgery (CABG). Cardiac nociceptor activation mechanisms were investigated in heart biopsies from these 22 CABG patients; reverse transcriptase polymerase chain reaction analysis (RT-PCR) was conducted for adenosine and bradykinin receptor mRNA.^9,10

An age-related decrease was observed in the adenosine A1 mRNA density but not in the bradykinin receptor mRNA levels. The adenosine A1 receptor density also correlated with pain characteristics reported in a questionnaire. Making use of semiquantitative RT-PCR, cardiac tissue substrates were assessed to determine the expression of adenosine A1 and bradykinin B1/2 receptor mRNA densities. The outcomes were associated with the quality of pain, age, gender, medication, and duration of disease.^9,10

For evaluation of pain characteristics, we used questionnaires and objective pain scores. We found that qualitative age-related alterations in angina perception correlated with the development of the more “strangling” component of angina at older age. This observation may be explained, in part, by a reduction in adenosine A1 receptor mRNA expression in the heart, since bradykinin B1/2 receptor densities remain the same.^9,10

Project 3: Nociception of cardiac pain in unrestrained rats

Having identified neural pathways, we studied neurons that were activated during electrical neuromodulation. ¹¹ In search of a putative mechanism of action of electrical neuromodulation, we hypothesized that neuromodulation affects processing of nociceptive information within the central nervous system (CNS). To characterize neural activity we used expression of both the immediate early gene c-fos and the “late gene” or stress protein known as heat shock protein 72 (HSP72). c-fos was used to identify structures in the CNS affected by spinal cord stimulation. HSP72 was applied to ascertain whether spinal cord stimulation might operate as a stressor.¹²

Animal experiments were conducted on unrestrained unanesthetized rats implanted with a permanent catheter in the pericardial space; acute cardiac pain was triggered in this space using capsaicin as the algogenic substance.¹³ The autonomic cardiovascular responses were recorded with implantable telemetric devices. Behavioral responses were recorded on videotapes taken from the same animals in which the involved cerebral structures were characterized by analyzing cerebral immediate early gene expression. Quantification of data makes it possible to study the effects of electrical neuromodulation and analgesic drugs on perception of cardiac pain. To apply electrical neuromodulation, two electrodes were positioned and sutured epidurally at the spinal cord of the rats. One electrode was fixated at spinal nerve C7 and the other at T2. Furthermore, we studied the effect of spinal cord stimulation on behavior. Three hours after stimulation, the rats were sacrificed and their brains and spinal cords were removed.

The treated group showed regional increased c-fos expression in a select group of regions of the limbic system—periaqueductal gray, paraventricular hypothalamic nucleus, paraventricular thalamic nucleus, central amygdala, agranular and dysgranular insular cortex, (peri)ambiguus, nucleus tractus solitarius, and spinal cord—involved in the processing of pain and cardiovascular regulation, among other functions. Moreover, in both treated rats and controls, HSP72 expression was found in the endothelium of the enthorhinal cortex, the amygdala, and the ventral hypothalamus, but not in the neurons. The treated animals were significantly more alert and active than were the controls.

Thus, the rat model we developed appears to be suitable for studying potential mechanisms through which neuromodulation may act. Moreover, neuromodulation affects c-fos expression in specific parts of the brain known to be involved in regulation of pain and emotions. HSP72 expression is limited to the endothelium of certain parts of the CNS, and thus physical stress effects were excluded as a potential mechanism of neuromodulation. Finally, our experimental model identified regions corresponding with regional cerebral blood flow changes during neurostimulation in patients.⁸

PROJECT 4: BIDIRECTIONAL HUMORAL AND NERVOUS HEART-BRAIN INTERACTIONS

With respect to the emotional component of angina, we thought to study alternative pathways of communication between the heart and the brain. This idea occurred as a consequence of observations that many patients who suffer serious cardiac events, such as CABG or myocardial infarction, are confronted with a period of emotional problems following these events. So, from our experimental projects, the question became relevant as to whether emotional alterations in behavior following a cardiac life event may be executed by a humoral pathway from the heart to the brain, since, vice versa, the brain controls the heart through both nervous and humoral pathways. In other words, is it feasible that both humoral and neural pathways are involved, bidirectionally, in interactions between the brain and the heart?

Cardiac disease, proinflammatory cytokines, and blood-brain barrier damage

Cardiac ischemia, the underlying cause of cardiac pain in angina pectoris, triggers a cascade of events that release numerous substances in the myocardium and circulation, all of which are potential candidates for nociceptor activation and initiation of behavioral and autonomic responses to cardiac pain. Some of the substances that are released into the circulation may play a role in the humoral communication between heart and brain, but when released chronically, these substances may induce neuropathological modifications. Anxiety disorders and depression are cerebral disorders that are frequently comorbid with ischemic heart diseases. The latter are attributed to noncoping behavior, but our own experiments (as part of the program) showed that immune activation after tissue damage in the heart generates regional blood-brain barrier damage (Project 4) that could be an underlying organic basis for comorbid neuropsychiatric disorders. The incentive for this project in general was the observation that myocardial infarction is accompanied by behavioral and neuronal abnormalities.

In this project we established whether release of proinflammatory cytokines after tissue damage in the heart is a possible inducer of comorbid neuropsychiatric diseases.

As a model for immune activation, we studied the effects of intravenous injections of the proinflammatory recombinant tumor necrosis factor–alpha (TNF-α) on cerebral endothelial leakage, induction of neuronal damage, and motor and cognitive function in rats. Determinants of selectivity of blood-brain barrier damage were assessed with a molecular biological approach in which we studied regional differences of TNF-α–induced expression in the cerebral endothelial cells of the immediate early gene c-fos and proteins involved in leukocyte docking (intercellular adhesion molecules [ICAMs]) and TNF-α receptors.

To examine the mechanisms by which this interaction occurs, we induced myocardial infarction in a group of rats and then performed immunohistochemistry of the brain. This experiment revealed regional serum protein extravasation, pointing to leakage of the blood-brain barrier. This process occurred in certain cortical, subcortical, and hindbrain areas in discrete patches. The leakage was colocalized with expression of the immune activation marker ICAM-1. To assess the involvement of the immune system in the effects shown, a second group of rats was injected with TNF-α, as the major proinflammatory cytokine. This procedure rendered the same results. It was concluded that myocardial infarction may interfere with the integrity of the blood-brain barrier and possibly with brain functioning through activation of the immune system. The relevance for pathophysiological processes may provide a substrate for further research in unraveling the emotional consequences of serious cardiac events.

In the state of immune activation that follows myocardial ischemic events, various cytokines are released from the myocardium into the plasma. These cyto kines potentiate the cytotoxicity of TNF-α. In the next experiment we were able to demonstrate that intravenous injection of TNF-α induces a selective and regional neural IgG and endothelial ICAM-1 immunoreactivity. The expression of TNF-α–induced changes in the brain suggests that TNF-α is capable of inducing blood-brain barrier dysfunction. It is hypothesized that through dysfunction of the blood-brain barrier, the released cytokines bind to specific cognitive centers in the brain and thus may lead to emotional disturbances following cardiac events.¹⁴

Having identified some specific centers involved in cardiovascular control, we further studied the effects of electrical and chemical stimulation of a specific brain center on the heart.

PROJECT 5: EFFECT OF BRAIN STIMULATION ON CORONARY FLOW

From a clinical PET study performed in patients with end-stage CAD during active spinal cord stimulation therapy, as well as from our PRV experiments and the literature, we concluded that the periaqueductal gray plays a central role in the regulation of different cardiovascular responses and in the integration of motor output from the limbic system.^6,7 Subsequently, the peri aqueductal gray has been thought to be one of the pivotal cerebral centers involved in executing electrical neuromodulation effects.

We investigated the function of the periaqueductal gray in regulation of the coronary flow of the heart. Depending on the stimulation site, electrical stimulation in the periaqueductal gray resulted in increases and decreases in coronary flow and conductance. These effects were organized topographically. The sites producing increases in coronary flow and conductance were found in both the dorsolateral and the ventrolateral periaqueductal gray. The sites producing decreases were restricted mainly to the ventrolateral portion. Similar topographic distributions were observed for the sites producing changes in carotid conductance and heart rate, but not for those producing changes in blood pressure and carotid flow. It is hypothesized that the topographic distribution of coronary vasoconstrictive and vasodilatory responses from the periaqueductal gray may enable optimal adjustments of the coronary perfusion. These optimal adjustments can then accommodate variations in myocardial oxygen demands accompanying different behavioral modes.

CONCLUSION

From all our experiments, mainly performed in rats (but sometimes also in a cat model due to the existence of a stereotactic brain atlas for the cat), we have learned about heart-brain communication through the use of electrical neuromodulation. In the last decade we have further studied heart-brain interactions in the International Working Group on Neurocardiology (IWGN), making use of canine and rabbit models. The main focus of the IWGN is on neural hierarchy in cardiac control. These projects are discussed by one of us (R.D.F.) elsewhere in these proceedings. In brief, the importance for the heart of the intracardiac neuron system and controlling centers at the C1 spinal level,^15–17 in conjunction with the induction of myocardial ischemia, will be highlighted. For a more extensive overview of recent work performed by the IWGN, see the reviews by Foreman et al¹⁸ and Wu et al.¹⁹